Studies in Systems, Decision and Control 65

Tian Seng Ng

Real Time
Control
Engineering
Systems and Automation
Studies in Systems, Decision and Control

Volume 65

Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: [email protected]
About this Series

The series “Studies in Systems, Decision and Control” (SSDC) covers both new
developments and advances, as well as the state of the art, in the various areas of
broadly perceived systems, decision making and control- quickly, up to date and
with a high quality. The intent is to cover the theory, applications, and perspectives
on the state of the art and future developments relevant to systems, decision
making, control, complex processes and related areas, as embedded in the fields of
engineering, computer science, physics, economics, social and life sciences, as well
as the paradigms and methodologies behind them. The series contains monographs,
textbooks, lecture notes and edited volumes in systems, decision making and
control spanning the areas of Cyber-Physical Systems, Autonomous Systems,
Sensor Networks, Control Systems, Energy Systems, Automotive Systems, Bio-
logical Systems, Vehicular Networking and Connected Vehicles, Aerospace Sys-
tems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power
Systems, Robotics, Social Systems, Economic Systems and other. Of particular
value to both the contributors and the readership are the short publication timeframe
and the world-wide distribution and exposure which enable both a wide and rapid
dissemination of research output.

More information about this series at http://www.springer.com/series/13304

Tian Seng Ng

Real Time Control

Engineering
Systems and Automation

123
Tian Seng Ng
School of Mechanical and Aerospace
Engineering
Nanyang Technological University (NTU)
Singapore
Singapore

ISSN 2198-4182 ISSN 2198-4190 (electronic)

Studies in Systems, Decision and Control
ISBN 978-981-10-1508-3 ISBN 978-981-10-1509-0 (eBook)
DOI 10.1007/978-981-10-1509-0

Library of Congress Control Number: 2016940881

© Springer Science+Business Media Singapore 2016

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained
herein or for any errors or omissions that may have been made.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer Science+Business Media Singapore Pte Ltd.
In Memory of
Dr. Lim Tau Meng
Preface

Control technology has been in existence for the last 60 years. Throughout the
decades, many developments have evolved in control and automation engineering.
Especially, nowadays, when computers have become more popular, technology in
control has progress to combine with computer technology for faster and more
precise method of computations. Software control can take the place of hardwired
control system economically. Hence, new control methods for real-time systems are
progressively discovered and taught in institutions. Most engineering applications
involve control function. Broadly speaking, there are many types of control tech-
nology in engineering domain. We have the electrical and electronics control
design, microcontroller and embedded system programmings and control, as well as
the mechatronics control system. The low-level assembly programming language
performs basic control techniques as well as controlling the stepper motor. Besides,
we can find control applications in big and complex industrial system. Power
system analysis can predict, monitor and therefore, control the load flow network
system. PLC system enhances the design of the elevator control system. We study
process control engineering to apply the theory to the water control system. More
advanced control technology such as the neural network machine learning tech-
nology finds its application in the chemical control plant. Furthermore, computer
vision technique is being used widely in the manufacturing factories or the
industries. As can be seen, the usefulness of real-time control engineering is applied
to countless industrial applications.
The book introduces many different types of control with relevances to real life
control systems design. Illustrative diagrams, circuits programming examples and
algorithms show the details of the system function design. Readers will find various
real-time control automation engineering practices and applications for the modern
industries as well as the educational sectors.

Singapore Tian Seng Ng

May 2016

vii
Acknowledgement

Thanks to the project supervisor, Dr. Fazlur Rahman, for his guidance on the topic:
control of process plant using the neural network. Gratefully thanks to Prof. Fok Sai
Cheong for his contribution to the control of electro-pneumatic mechanisms. Also,
appreciation goes to the technical staff Mr. Lim Chai Lye for assisting in the
servo motor control system modification. Besides, the author is also thankful to
Mr. Ang Kwee Leng, the engineer for giving his unassuming support on the
electrical system.
It was sad that the associate professor, Dr. Lim Tau Meng, had left us. His past
contribution to the mechatronics section is highly appreciated. The author would
also like to thank Prof. Yap Kian Tiong for his teachings and advises on micro-
controller system and Mr. Norbel Navarro for some of the microcontroller pro-
grammings. Not forgetting the support team assisting Prof. Yap on the
microcontroller section. They are Mdm Yap-Lee Koon Fong, Mrs. Pamela Loh,
Mrs. Grace Ho, Mrs. Josephine Loh, Mr. Seow Tzer Fook and Mr. Ng Jui Hock
helping in the engineering laboratory.
Appreciation is due to my family members and friends for their overall support
and patience. Without these people, the writing of the book would not have been a
success. Finally, I would like to thank NTU, School of Mechanical Engineering,
Division of Mechatronics and Design, for giving me the opportunity to make this
work a reality.

Singapore Tian Seng Ng

May 2016

ix
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Highlights of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Organisation of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Embedded Intruder System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Requirements and Assumptions . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Hardware Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Software Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 System Program. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3 Mechatronics . . . . . . . . . . . . . . . . . . ....... . . . . . . . . . . . . . . . 27
3.1 Liquid Level Control . . . . . . . . ....... . . . . . . . . . . . . . . . 27
3.2 Oscillating Planar. . . . . . . . . . . ....... . . . . . . . . . . . . . . . 29
3.3 Conveyor Inspection Using Shift Registers . . . . . . . . . . . . . . . 34
3.4 Modern Speed Control . . . . . . . ....... . . . . . . . . . . . . . . . 35
4 Microcontroller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.1 Basic I/O Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 LCD and Keypad. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.3 Waveform Timings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.4 Pressure Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.5 Temperature Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.6 Stepper Motor Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.7 Serial Communications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5 Electronics Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.1 Servo Motor Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.2 Square Wave Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.3 PID Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.4 Control of an Electro-pneumatic Mechanism . . . . . . . . . . . . . . 87

xi
xii Contents

6 Electrical System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.1 Elevator Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.2 Programmable Logic Controller . . . . . . . . . . . . . . . . . . . . . . . 91
6.3 Ladder Diagram Control Structures . . . . . . . . . . . . . . . . . . . . 91
6.3.1 Part 1:- Indicating Lights . . . . . . . . . . . . . . . . . . . . . 93
6.3.2 Part 2:- Lift Door Open/Close . . . . . . . . . . . . . . . . . . 94
6.3.3 Part 3:- Lift Up/Down . . . . . . . . . . . . . . . . . . . . . . . 96
6.4 Safety Control Features. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7 Power Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.1 Power System Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.2 Newton Raphson Formulation . . . . . . . . . . . . . . . . . . . . . . . . 101
7.3 Load Flow Analysis Using Newton Raphson. . . . . . . . . . . . . . 104
8 Process Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
8.1 Water Tank Control System . . . . . . . . . . . . . . . . . . . . . . . . . 115
8.1.1 First-Order Derivation. . . . . . . . . . . . . . . . . . . . . . . . 115
8.2 Single Tank Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
9 Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
9.1 Neural Network in Process Control . . . . . . . . . . . . . . . . . . . . 121
9.2 The Artificial Neurons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
9.3 Techniques Involved in the Controllers . . . . . . . . . . . . . . . . . . 123
9.4 NN Learning Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
9.5 Selection of the Learning Algorithms . . . . . . . . . . . . . . . . . . . 126
9.6 The Network Topology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
9.7 MLP Backpropagation Network for Process Control. . . . . . . . . 130
9.8 Chemical Plant NN Feedback Control System . . . . . . . . . . . . . 133
9.8.1 Process Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
9.8.2 Process Verification . . . . . . . . . . . . . . . . . . . . . . . . . 135
9.8.3 Process Improvement . . . . . . . . . . . . . . . . . . . . . . . . 137
9.9 Remote Operated Neural Network Control Plant . . . . . . . . . . . 138
9.9.1 Field Instrumentations. . . . . . . . . . . . . . . . . . . . . . . . 138
9.9.2 Scaling and Conversions . . . . . . . . . . . . . . . . . . . . . . 139
9.9.3 Control Valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
9.9.4 Wireless Transmissions . . . . . . . . . . . . . . . . . . . . . . . 140
9.10 Valves and Chemical Plant Tunings . . . . . . . . . . . . . . . . . . . . 142
9.10.1 Desired Chemical Mixture, Samples and NN Data . . . . 142
9.10.2 Chemical and Valves Calibration . . . . . . . . . . . . . . . . 142
9.10.3 Trial Test in Actual Plant . . . . . . . . . . . . . . . . . . . . . 143
9.11 Computerized Neural Network Control System . . . . . . . . . . . . 144
9.11.1 NN Real Time Control Plant . . . . . . . . . . . . . . . . . . . 144
9.11.2 Neural Network Control Valves . . . . . . . . . . . . . . . . . 145
9.11.3 Intelligent Advisor . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Contents xiii

10 Computer Vision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

10.1 Image Thresholding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
10.2 Zhang-Suen Thinning Algorithm . . . . . . . . . . . . . . . . . . . . . . 154
10.3 Brief Descriptions of the Program Algorithms . . . . . . . . . . . . . 154
10.4 Image Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

Appendix A: MC68HC11 Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

Appendix B: MCU Port Testers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

Appendix C: LCD References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
List of Figures

Figure 2.1 Hardware layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Figure 2.2 Memory and I/O map. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Figure 2.3 Microprocessor interface . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Figure 2.4 RAM decoding circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Figure 2.5 EPROM decoding circuit . . . . . . . . . . . . . . . . . . . . . . . . . 12
Figure 2.6 Alternative solution to RAM/ROM decoding. . . . . . . . . . . . 13
Figure 2.7 Input device interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Figure 2.8 Output device interface. . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Figure 2.9 Providing interrupt type to interrupt request . . . . . . . . . . . . 17
Figure 3.1 Transistor TIP-121 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Figure 3.2 Latching circuit for filling water tank . . . . . . . . . . . . . . . . . 30
Figure 3.3 Microcontroller controlled latching circuit . . . . . . . . . . . . . . 31
Figure 3.4 L298 H-bridge dual bidirectional motor driver . . . . . . . . . . . 31
Figure 3.5 Bidirectional DC motor control . . . . . . . . . . . . . . . . . . . . . 32
Figure 3.6 Oscillating planar circuit . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Figure 3.7 Conveyor system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Figure 3.8 Shift register conveyor control. . . . . . . . . . . . . . . . . . . . . . 36
Figure 3.9 PWM control of a small permanent magnet motor . . . . . . . . 37
Figure 3.10 RC differentiator pulse signal . . . . . . . . . . . . . . . . . . . . . . 38
Figure 4.1 8-bit dip switch input circuit . . . . . . . . . . . . . . . . . . . . . . . 40
Figure 4.2 LEDs output circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Figure 4.3 Buzzer and push button circuit . . . . . . . . . . . . . . . . . . . . . 42
Figure 4.4 A/D converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Figure 4.5 Keypad and LCD using keypad encoder . . . . . . . . . . . . . . . 50
Figure 4.6 Keypad and LCD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Figure 4.7 PWM control of DC motor . . . . . . . . . . . . . . . . . . . . . . . . 62
Figure 4.8 Pressure sensing circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Figure 4.9 Temperature sensing circuit. . . . . . . . . . . . . . . . . . . . . . . . 68
Figure 4.10 Stepper motor driving circuit . . . . . . . . . . . . . . . . . . . . . . . 69
Figure 5.1 Servomotor controller box. . . . . . . . . . . . . . . . . . . . . . . . . 80

xv
xvi List of Figures

Figure 5.2Error detecting circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

Figure 5.3Servomotor control block diagram . . . . . . . . . . . . . . . . . . . 81
Figure 5.4Square wave oscillator circuit . . . . . . . . . . . . . . . . . . . . . . 82
Figure 5.5Timer 555 square wave generator . . . . . . . . . . . . . . . . . . . 83
Figure 5.6Differentiator, integrator, proportional circuits . . . . . . . . . . . 84
Figure 5.7PID analogue controller . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Figure 5.8Frequency response bode plot . . . . . . . . . . . . . . . . . . . . . . 86
Figure 5.9Amplifier circuits a Differential. b Invertor. c Summing . . . . 87
Figure 5.10
Analogue signal control circuit . . . . . . . . . . . . . . . . . . . . . 88
Figure 5.11
Static sensitivity graph . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Figure 6.1Lift references . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Figure 6.2Indicating lights. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Figure 6.3Indicating lights alternate design . . . . . . . . . . . . . . . . . . . . 94
Figure 6.4Lift open/close . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Figure 6.5Lift up/down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Figure 6.6Lift safety control features . . . . . . . . . . . . . . . . . . . . . . . . 99
Figure 7.1Result of the N-R computation . . . . . . . . . . . . . . . . . . . . . 103
Figure 7.2Two-bus power network circuit . . . . . . . . . . . . . . . . . . . . . 105
Figure 7.3Transmission line admittance circuit . . . . . . . . . . . . . . . . . . 106
Figure 7.4DOS environment (Y-parameters input) . . . . . . . . . . . . . . . 106
Figure 7.5Case 1 output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Figure 7.6Case 2 output (without inductor) . . . . . . . . . . . . . . . . . . . . 107
Figure 7.7a Case 2 output, b case 2 output, c case 2 output,
d case 2 output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Figure 7.8 Case 3 output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Figure 8.1 Single tank diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Figure 8.2 System block diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Figure 8.3 Water tank apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Figure 8.4 RC time constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Figure 8.5 Proportional alone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Figure 8.6 PI output (fixed proportional) . . . . . . . . . . . . . . . . . . . . . . 119
Figure 8.7 PI output (fixed integral). . . . . . . . . . . . . . . . . . . . . . . . . . 119
Figure 9.1 Neural network process control . . . . . . . . . . . . . . . . . . . . . 122
Figure 9.2 Neural network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Figure 9.3 Activation graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Figure 9.4 Hebb learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Figure 9.5 Perceptron learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Figure 9.6 Delta learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Figure 9.7 Perceptron network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Figure 9.8 Backpropagation network . . . . . . . . . . . . . . . . . . . . . . . . . 127
Figure 9.9 Best generalisation graph . . . . . . . . . . . . . . . . . . . . . . . . . 128
Figure 9.10 Mixer plant neural net control system. . . . . . . . . . . . . . . . . 130
Figure 9.11 Neural net system processor . . . . . . . . . . . . . . . . . . . . . . . 131
Figure 9.12 Neural network external structure. . . . . . . . . . . . . . . . . . . . 131
List of Figures xvii

Figure 9.13 NN output neurons (with secondary axis) . . . . . . . . . . . . . . 136

Figure 9.14 Neural network output layer . . . . . . . . . . . . . . . . . . . . . . . 136
Figure 9.15 Auto adjustment trends of the NN system . . . . . . . . . . . . . . 138
Figure 9.16 IEC 62591 interfacing network . . . . . . . . . . . . . . . . . . . . . 139
Figure 9.17 On/off valve control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
Figure 9.18 Wireless valve automation. Courtesy
of Emerson Process Management. . . . . . . . . . . . . . . . . . . . 141
Figure 9.19 Automatic neural network control system . . . . . . . . . . . . . . 141
Figure 9.20 Chemical plant valves and sensors . . . . . . . . . . . . . . . . . . . 143
Figure 9.21 Two heat exchanger units . . . . . . . . . . . . . . . . . . . . . . . . . 144
Figure 9.22 Control valves operating ranges . . . . . . . . . . . . . . . . . . . . . 144
Figure 9.23 Neural network I/O block . . . . . . . . . . . . . . . . . . . . . . . . . 146
Figure 9.24 NN training result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Figure 9.25 Simulated result. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Figure 9.26 Quality matching advisor . . . . . . . . . . . . . . . . . . . . . . . . . 149
Figure 10.1 Chinese characters (images sets 1 and 2) . . . . . . . . . . . . . . 161
Figure 10.2 More Chinese characters (images sets 3–5) . . . . . . . . . . . . . 162
Figure 10.3 Hao Chinese character (image set 6). . . . . . . . . . . . . . . . . . 163
Figure 10.4 Scanned original character (image set 7) . . . . . . . . . . . . . . . 164
List of Tables

Table 2.1 Decoding table for Y11 output . . . . . . . . . . . . . . . . . . . . . . 14

Table 3.1 L298 logic control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Table 6.1 Elevator I/O functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Table 6.2 Elevator on-delay timer settings . . . . . . . . . . . . . . . . . . . . . . 100
Table 9.1 Weights tabulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Table A.1 Control and display command . . . . . . . . . . . . . . . . . . . . . . . 173
Table A.2 The command control codes . . . . . . . . . . . . . . . . . . . . . . . . 174

xix
List of Flow Charts

Flowchart 2.1 Security system flow chart . . . . . . . . . . . . . . . . . . . . . . . 18

Flowchart 4.1 LCD display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Flowchart 4.2 Keypad and LCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

xxi
List of Programs

Program 2.1 Microprocessor Security System . . . . . . . . . . . . . . . . . . . 20

Program 3.1 Liquid Level Control. . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Program 3.2 Oscillating Planar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Program 4.1 Read Dip Switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Program 4.2 Write to Output LEDs . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Program 4.3 Push Button & Buzzer. . . . . . . . . . . . . . . . . . . . . . . . . . 41
Program 4.4 A/D Converter 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Program 4.5 A/D Converter 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Program 4.6 A/D Converter 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Program 4.7 A/D Converter 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Program 4.8 LCD Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Program 4.9 Keypad & LCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Program 4.10 Keypad & LCD In C Program . . . . . . . . . . . . . . . . . . . . 56
Program 4.11 Square Wave Timer . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Program 4.12 PWM Signal of 400Hz at 2.5ms Period . . . . . . . . . . . . . . 60
Program 4.13 PWM Motor Control With A Potentiometer . . . . . . . . . . . 63
Program 4.14 Two Phase Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Program 4.15 Full Step Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Program 4.16 Half Step Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Program 4.17 PC Communication With MicroP . . . . . . . . . . . . . . . . . . 77
Program 7.1 Newton Raphson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Program 7.2 Newton Raphson Load Flow . . . . . . . . . . . . . . . . . . . . . 111
Program 9.1 Neurons Simulated Valves . . . . . . . . . . . . . . . . . . . . . . . 147
Program 9.2 Computerized NN Quality Advisor . . . . . . . . . . . . . . . . . 150
Program 10.1 Binarise Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Program 10.2 Thinning Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

xxiii
Chapter 1
Introduction

1.1 Objectives

Computer controlled automation, are becoming very common as we apply it in

daily life. Advances in technological improvements have made it find its applica-
tions in automation. Therefore, the study and understanding of the scope are nec-
essary for building real-time automation systems. More complicated automatic
systems require a thorough grasping of the related topics in the book. The materials
presented in this book projects the modern control systems and its applications in
real time. Readers can gain knowledge in the field of electronics and microcon-
trollers. The domain provides a good foundation for laboratory-based experiments
and practices. Moreover, the readers can also appreciate the electrical knowledge
and its applications presented here. Excerpts of the load flow network open an
insight to electrical engineers. Interesting computer and programmable control
techniques can enhance the interest of the book explorers. We can discover new
topics like the neural network automated system and modern technology like
computer vision in the book.

1.2 Highlights of the Book

The book introduces many different kinds of control systems and design. They
comprise of hardware and software for the real-time engineering control systems.
• Security alarm system is commonly being practiced in almost every organisa-
tions. It provides intruder detection during after office hours. We introduced the
study of the microprocessor controlled alarm system.
• Electronic circuits were built to accommodate each of the different design
structure. We construct various kinds of the mechatronics systems. Some are

© Springer Science+Business Media Singapore 2016 1

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_1
2 1 Introduction

programmed using the PBasic programming language. In the book, different

types of mechatronics hardware are introduced and appreciated.
• We integrate the microcontroller to perform control and measurements.
Numerous types of software are programmed in assembly language to control
the hardwired system. We used the 68HC microcontroller family in the system
design.
• The design, build and study of the servo motor is found in the book. We
experimented the electronic circuitries and analysed the characteristics of the
servo motor.
• Programming Logic Control applies in the elevator control design. Ladder
diagrams drawn show the flexibility and complexity in the development
configuration.
• The complexities of the power system analyses for large system network is
being presented in the following chapter. We used the Pascal language in the
computations for the power flow calculations.
• Applied control theory is presented to determine the water process control
system. Laboratory water tank experiment accelerates the control application.
• The book narrates the control of chemical plant using the neural network for the
up to date development in control technology and applications.
• Very often, we applied computer vision technique to the modern industry
control system for scanning the defects in the production lines.

1.3 Organisation of the Book

• Chapter 2 analyses the design of the microprocessor security system. We

introduce and discuss the hardware circuits and software program in assembly
language.
• Chapter 3 describes the many different types of mechatronics control systems.
We include diagrams and circuits for the design of the mechatronics systems for
better understandings.
• Chapter 4 presents the microcontroller and its hardware applications. We pro-
grammed to perform basic i/o functions. Signal conditioning circuits for tem-
perature and pressure measurements are shown. Three different stepper motor
control techniques are analysed.
• Chapter 5 introduces the control of a servo motor. Square wave generator is
developed for the control input into the servo motor. We present the PID
controller in the system. Besides, we also use an analogue controller to control
an electro-pneumatic mechanism.
• Chapter 6 highlights the application of the lift control system by using logic
ladder diagram. The lift up/down, open/close and light indications functions are
illustrated in the complete elevator system.
1.3 Organisation of the Book 3

• Chapter 7 illustrates the power flow techniques for network system analysis.
Programming examples in Pascal describes a clear view of how to monitor the
load flow system.
• Chapter 8 study the control characteristics of the water tank control system.
Experimental set characterises the single tank control system.
• Chapter 9 extends the plant process control by using neural networks. The topic
includes the use of the machine learning technique to control the liquid flow of a
chemical process plant by controlling the valves.
• Chapter 10 illustrates the use of computer vision. The developed algorithm
skeletonized the main outlines of the picture for control and application
purposes.
Chapter 2
Embedded Intruder System

A microprocessor is a powerful device used to control the input and output oper-
ations of an external device. It serves as a data storage element as well. There are
various stages of upgrade and development of the 8086 (16-bit) microprocessor
unit. Its applications are used widely in the commercialized and industrialised areas.
For example, it is used to perform arithmetic and logical operations inside of a
computer system. It enables high speed and large memory storage in computer
application, especially in the latest development of the computer system. Besides,
CISCO securities also incorporate microprocessor in their security systems.
A standard microprocessor can be programmed and interfaced to the external
devices to control and operate a remote control system or collect and perform data
operations. An intruder alarm system safeguards an organisation’s asset and
security. To implement the automation, security system, we need to look at the
drawing of the building floor plan and layout. Both the hardware and software
combine for the intruder system to work.
The chapter comprises of the detailed design of the microprocessor hardware as
well as software algorithms of a security system in the computer room. The scripts
described the system analysis, specifications, selection and design of the security
system. It also includes hardware interface design and software flow charts and
programs. The valuable experience gained and knowledge applied were briefly
described in the following few pages.

© Springer Science+Business Media Singapore 2016 5

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_2
6 2 Embedded Intruder System

2.1 Requirements and Assumptions

Hardware Requirements:
• 8086 CPU or its upgraded version
• Microprocessor clock signal of 5 MHz
• 8 KB EEPROM
• 4 KB RAM
• Switches: 5 numbers excluding the reset switch
– switch 1, 2, 3 for front, side and back door sensors
– switch 4 to turn on autotime switch
– switch 5 for external interrupt to alarm
• 7 segment LEDs (8 nos)
– 6 LEDs to display time in hrs, mins and secs
– 2 LEDs to display 30 s countdown.
System Analysis:
• At power up system reset and display blank until sw4 is activated.
• Upon pressing the switch, the system starts up and corresponding time of the
day is displayed.
• Security system starts from 19:00:00 pm to 07:00:00 am.
• System constantly checked the current time with the limit time.
• If matched, it checks the memory location (60020H) for data.
• After activation, only 30 s are allowed to enter the correct code.
• The system will trigger the alarm if 30 s expire without the correct code entered.
• The system will not trigger an alarm if the correct code is being entered within
the time.
• Data will be stored in the memory address (60020H) if someone had entered
between the time.
• Otherwise, no data in the location.
Stated Assumptions:
Software:
• Clock generator had no wait state.
• Delay of counting sequence omitted in the program.
• Six secret codes to enter including alphabetical letters within 30 s.
• Assumed program time is aligned to the real-time clock.
2.1 Requirements and Assumptions 7

Hardware:
• No buffer used for the output device as it is being connected to less than ten
output.
• The time colons is by other physical means without using 7-segments.
• External hardware interrupt service provided via interrupt type 60H.
• A reset button (normally closed) is physically connected to the microprocessor
via the supply line.
• 4 KB RAM excluding interrupt address.
• Interrupt vector uses another 1 KB RAM.
• 500 byte RAM piece exist.

2.2 Hardware Design

The internal architecture of the CPU consisted of the bus interface unit and the
execution unit. They performed the fetch, decode and execution operations. The
microprocessor CPU registers enhanced speed processing of data. They are mainly
the general purpose registers; pointers and index registers; segment registers and
flag registers. The 8086 is used to operate in the minimum mode for single CPU
environment in the system. 5 MHz clock operation is designed using 8254A chip
and interfaced into the microprocessor. The crystal of the 8254A clock generator is
three times the microprocessor input clock. Alternatively, you can use 8284A clock
generator to generate a direct 5 MHz timer for the 8086/8088 microprocessor [20].
The selection of the memory or input/output operation enables 1 MB of memory
addresses and 64 KB of i/o addresses. The address lines share with the 16 bits data
lines. We use latches for latching the address lines in addition to another four
address lines. Making a total of 20 address lines. Therefore, it saves the spaces of
the microprocessor. Multiplexers are used in the address line so data will not go to
the wrong location when we change the address. At the other end, it is demulti-
plexed. The decoding circuits are used to enable one selection of the addresses at a
time. The memory used in the design is 24 KB or 8 KB EEPROM and 2 * 2 KB or
4 KB SRAM. We select static RAM for faster operation due to its non-volatile
characteristics. Moreover, we take also the advantages of eliminating the refreshing
circuit. The switches connect to the buffer 74LS244 for boosting the signal to the
microprocessor. Where latches, 74LS374 are being connected to the seven segment
LEDs. The purpose is for stabilizing the output ports. We described the selected
design with diagrams (see Figs. 2.1, 2.2 and 2.3).
8 2 Embedded Intruder System

Fig. 2.1 Hardware layout

 2.2 Hardware Design

Fig. 2.2 Memory and I/O map

9
 10
2 Embedded Intruder System

Fig. 2.3 Microprocessor interface

 2.2 Hardware Design

Fig. 2.4 RAM decoding circuit

11
12 2 Embedded Intruder System

EPROM Address FE000H FFFFFH

A19–A16 A15A14A13A12 A11-A8 A7-A4 A3-A0
FH 1 1 1 X XH XH XH

VCC

74LS10

Fig. 2.5 EPROM decoding circuit

In the RAM decoding circuit of Fig. 2.4, the read ðRDÞ and the write ðWRÞ
operations enable the selections of the data read, and data write operations
respectively. We used logic gates to perform the functions of the combinational
logics for the addresses. Thus, RAM addresses from 60000H to 60FFFH in the
memory map are selectable.
In the EPROM decoding circuit, 2 nos of 74LS08 are required for the 7  2
input AND gates to be joined up for the eight inputs. A0 address forms the nec-
essary selection for the odd or even bank selection (see Fig. 2.5).
Next, we have an alternative solution to the RAM/ROM decoding circuit. The
stated assumption in the scenario is that the 1.5 KB and 0.5 KB RAM exists.
The circuit in the top left corner in Fig. 2.6 is derived from the following table.
The decoding map represents the 3 rows of memory addresses (see Table 2.1).
Following up is the input decoding device interfacing. The addresses for the
input devices range from 0060H to 007EH. The dip switches selections serve as the
data as well as the address lines input to the microprocessor unit. The addresses A0
to A4 are selectable by the dip switch input. The relevant dip switch selects the
active low for each of the matching input address. For example, each of the dip
switches represents each sensor. When any of the five sensors activates, a logic, ‘0’
will select the corresponding address input. So data for the selected address is input.
NOR gate helps to decode the address lines A9 to A12. Three of the dip switches
are replaceable by door sensors. The input keypad circuit can be designed from
address 0071H to 0076H (0000 0000 0111 0xxx) also. Alternatively, we can use
8279 chip for interfacing the keypad to the microprocessor (see Fig. 2.7).
 2.2 Hardware Design

Fig. 2.6 Alternative solution to RAM/ROM decoding

13
14 2 Embedded Intruder System

Table 2.1 Decoding table for Y11 output

The output device as shown in Fig. 2.8 is interfaced to the 8 nos. of 7 segment
LED display. The seven segment LEDs are for hours, minutes, seconds and the 30 s
countdown as stated in the design assumption. We used the BCD to 7 segment
driver (74LS47) to light the LEDs. Thus, only the first 4-bit data is enough to
display the ten different numbers. The 74LS138 decoder is used to decode the
selectable output for display. The decoded output addresses are from 0050H to
0057H. The interrupt type 60H is input into the microprocessor once there is an
interrupt request for calling the interrupt subroutine. We can use input port A4 of
Fig. 2.7 to call the interrupt request as in Fig. 2.9. The microprocessor will
acknowledge the interrupt request to accept the interrupt type.

2.3 Software Design

We include a flow chart, (see Flowchart 2.1) of the program which consists of the
following functions:
(a) The auto time will start only when switch 4 activates.
(b) The program consistently scans for the time between 19:00:00 and 07:00:00
interval.
(c) Next, if it is within the time interval, a subroutine will check for the external
interrupts. In another word, we scan switch 1 (front door), switch 2 (back
door) or switch 3 (side door) for anyone entering the room. If switches are not
activated, it will loop back and forth the main program and the subroutine until
the time is not within the time interval.
 2.3 Software Design

Fig. 2.7 Input device interface. Note Dip sw5 activates the input port number 006FH and so on
15
 16

74LS47

0050H

0051H
0052H
A4
A6 0053H
0054H
0055H
0056H
A3 0057H
A5
A7

A15

0050H 0057H

0 0101

Fig. 2.8 Output device interface

2 Embedded Intruder System
2.3 Software Design 17

Fig. 2.9 Providing interrupt type to interrupt request

(d) If any one of the switches is activated, it does two things. First, it activates the
30 s countdown. At the same time, it will also check for the correct code
sequence entered. The program will check for one input code each time one
second passes, by going through the delay procedure. It will compare with the
correct code sequence of the six secret codes. Therefore, the program allows
the intruder to key in a total of 30 code words before it sounds the alarm.
Normally, to get the correct security password take 6 s to key in. But if the
user key in wrongly, he may take more than 6 s to do it. If the system detects
the correct code within 30 s, it will drop off the alarm.
The range between the first and the last wrong code entered is 54–74
instructions (only consumes 59.2 ls) within the program. If a second passes
and no code input, it will take the empty code and compare with the correct
code. Since it does not match, the outcome is that it will loop to re-match the
code again. Thereby, it lost one second. Similarly, if any of the subsequence
code does not match it will also losses a second.
A 16-bit loop will consist of approximately 65,600 maximum instructions.
Calculations for each second delay takes about 65,540 instructions per
loop. The program consumes a total of 0.052 s per loop (the calculated values
is: 4 machine cycles  200 ns per instruction, for 65,600 instructions).
Therefore, the program scans 19 loops for an entered code within a second. So
overall, the program may check between 28 and 30 times each time a code is
entered, to test for the correct secret code sequence within the 30 s.
18 2 Embedded Intruder System

Flowchart 2.1 Security

system flow chart
2.3 Software Design 19

(e) If any of the door sensors is activated and left opened while the code sequence
enter is correct, the program will continue to run auto time restarting the 30 s
countdown, to scan for new input codes to be entered. If it senses no input
entered, it will still sound the alarm if 30 s is up. So, all the doors should be
shut immediately after entering, between 19.00 pm and 07.00 am. Another
scenario is, if any of the doors are open and close after entering the correct
codes, the person will still have to re-enter the codes again.
(f) All the codes entered within the 30 s will be saved.
(g) If 30 s expired before we enter the correct code sequence, the subroutine
triggers an external interrupt to an alarm.
(h) All the above scenarios will activate the alarm accordingly unless the power
switch is turned off, or the switch 4 is not on. The alarm is turn off using the
reset button. The power switch, the reset button and switch 4 is highly secured.

2.4 System Program

We must incorporate hardware with software to run the system properly. An

advantage of the built-in system is that the alarm connected to an interrupt sub-
routine is triggered externally (see Fig. 2.9) by the hardware (switch 5) for testing
of the alarm system. Secondly, the six code sequence consisting of digits together
with alphabets, which can be displayed by the 7-segment) entered instead of four
digits provides a better security to the system. The stored code entered by the door
breaker can be retrieved by computerized mean by an interface to the micropro-
cessor. Thirdly, the alarm trigger number is stored in the memory location. It serves
to check for intruder break in for the case where the alarm fails to sound.
A disadvantage of the system is that it affects the auto timing when the 30 s
activates and the time taken to enter the code sequence. Thereby, causing a delay in
the actual timing of the system. It is considered as pros, as the delay in the timing
system might be a second alternative to hint for a broke-in between the time
interval, besides checking the code sequence stored. Later the system can be reset
back to the precise auto time again by the personnel, which have the key to the reset
button panel. We can also improve the system further.
20 2 Embedded Intruder System
2.4 System Program 21
22 2 Embedded Intruder System
2.4 System Program 23
24 2 Embedded Intruder System
2.4 System Program 25
26 2 Embedded Intruder System
Chapter 3
Mechatronics

Mechatronics systems are widely utilized in the automation, industries. The prac-
tices involve the controls of electronics for mechanical systems. In this chapter, we
present the liquid level control system, the oscillating planar mechanical system, the
conveyor inspection using shift registers and the modern speed control unit.

3.1 Liquid Level Control

The latching circuit for filling the water into a tank or beaker is as shown in
Fig. 3.2. We can conduct a simple experiment by using a solenoid valve to open
and close an opening for controlling the flow of water into a beaker. The operation
is simply a latched circuit for water flowing to a beaker. The apparatus requires a
water carrying container hooked onto a standing rod structure and a beaker. The bar
structure only has a square base and an upright standing bar or rod. The square base
holds the weight of the water container. We placed a beaker underneath the water
carrying container, on the base frame structure. A solenoid is attached below to the
container to open or close the valve. The solenoid valve is at normally closed when
not energise, to stop the water from flowing out of the container. A small pipe fixed
downward from the container leads the water to fill up the beaker.
The sensors required are a proximity sensor, a latch button and a spring button.
The latch button ‘S2’ is latched for the operation of the system. The operation starts
by pressing the ‘LS1’ spring button. The output of the ‘OR’ gate is activated.
The LED is lighted up by the logic ‘1’ from the ‘OR’ gate output. The signal also

© Springer Science+Business Media Singapore 2016 27

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_3
28 3 Mechatronics

Fig. 3.1 Transistor TIP-121

enters the buffer to energise the solenoid valve. Thus, the valve will open and water
start to flow out from the container into the beaker. The output of the ‘OR’ gate also
goes into the Pin 1 of the ‘AND’ gate. So the circuit is latched as the ‘AND’ gate
outputs a logic ‘1’ signal. In the circuit, the 1 kΩ resistors are being used as the
pull-down resistors. We connect the 1.8 kΩ resistors to the transistor base. The
470 Ω resistor serves as the pull-up resistor for the proximity sensor.
We placed the proximity switch at a certain level of the beaker. So that when the
water level reaches the level, the proximity switch will be activated to stop the
water flow. The ‘TIP-121’ transistor will be de-activated and a logic ‘0’ will go to
the pin 2 of the ‘AND’ gate. Thus, the output of the ‘AND’ gate will be zero. The
output LED will not lit up, and the solenoid valve will be de-energised. So the
opening of the valve will close back to stop the water from flowing out of the pipe.
If the latched button is de-latched during the water filling operation, it will stop the
water filling process. To continue the operation, we had to reset the latch switch to
the connected position. Then we pressed the ‘LS1’ spring button to continue the
water filling process (Fig. 3.1).
An illustration of a simple program written in the PBasic language is as shown
in the program P3.1. We make use of the Basic Stamp microcontroller to perform
similar function. The microcontroller connects to three input ports 5, 6 and 7 for
sensings. Input port 7 connects to the proximity sensor. While we connect the
input ports 5 and 6 to LS2 and S2 (latched) switches respectively. If both buttons
are turned on, it will activate the solenoid to open the valve to fill the beaker or
water tank. The proximity sensor (to IN7) fixed at a certain level of the tank must
also be cleared to begin the water flowing process. The output port 10 is con-
nected to activate the solenoid valve via a buffer and a TIP-121 transistor.
Similarly, a 1.8 kΩ resistor is connected to drive the base of the transistor. The
proximity sensor is activated when the water level reaches its height. Thereby
3.1 Liquid Level Control 29

--------------------------------------------
Program 3.1: Liquid Level Control
--------------------------------------------
'{$STAMP BS2}
DIRS=%0000000010000000
again IF IN7=1 AND IN5=1 THEN do
you:OUT10=0 ‘ water stop
GOTO again
do: IF IN6=0 OR IN7=0 THEN you
OUT10=1 ‘ water flow
GOTO do

causing IN7 to input a zero, resulting output port 10 to output a logic low to the
transistor. It will deactivate the solenoid, and close the valve to stop the water
from flowing out. By making use of the program, we can replace the hardware
circuits (Fig. 3.3).

3.2 Oscillating Planar

In this section, we combined the field of electronics with mechanical to build and
control an oscillating planar. The operation begins with the preset switch pressed
when powered up. The first D-latch on top presets with a logic ‘0’ and the second
D-latch below clears with a logic ‘0’. So the oscillating planar starts to turn to the
right. As the planar rotates, the slotted optical switch will be blocked and activates
again. It triggers clock input to the D-latches. As a result, a logic ‘0’ now output
from the top D-latch. The logic ‘1’ now appear at the output of the lower D-latch. It
reverses the turning direction of the oscillating planar. It will again activate both the
optical switches fixed on both sides of the planar’s edge. A second trigger to the
D-latches will stop the planar. Now both the outputs of the D-latches are zero.
Therefore, we can see that the function of the preset switch is to reset the position of
the oscillating planar to its starting position. The circuit is operated by a start switch
to hold a logic ‘1’ to pin 2 of the ‘AND’ gate. We begin the oscillating function by
pressing the preset switch. It triggers a logic high and a logic low from the output of
 30
3 Mechatronics

Fig. 3.2 Latching circuit for filling water tank

3.2 Oscillating Planar 31

330R

Fig. 3.3 Microcontroller controlled latching circuit

Fig. 3.4 L298 H-bridge dual bidirectional motor driver

32 3 Mechatronics

Table 3.1 L298 logic Input Function

control
VA = H C = H; D = L Turn right
C = L; D = H Turn left
C = D Fast motor stop
VA = L C = X; D = C Free running motor stop

Fig. 3.5 Bidirectional DC motor control

the D-latches to turn the planar. As the slotted optical switches senses, a clock
trigger to the D-latches will output a logic ‘0’ to the ‘C’ input of the motor. At the
same time logic ‘1’ enters the ‘D’ input of the same motor simultaneously. It
changes the direction of the moving planar. The oscillating planar reverses direction
again when the optical switches activate a second time. The operation continues
until the ‘start’ holding switch stops. So we can see that we use the same button for
the start and the stop operation (Fig. 3.4 and Table 3.1).
The 4 diodes used for the DC motor are 1 A high speed diodes (Figs. 3.5, 3.6).
 IN4001

gnd
3.2 Oscillating Planar

74LS74A

0.5R / 1w

Fig. 3.6 Oscillating planar circuit

33
34 3 Mechatronics

----------------------------------------
Program 3.2: Oscillating Planar
----------------------------------------

'Oscillating Table Solenoid Sense:Software to replace flip-flop control

'in2=clear; in5=reject; in6=sensor; out12=red led
'out13=yellow led; out14=green led; out15=solenoid+orange led
DIRS=%1111111100000000
cnt VAR Byte
gone: cnt=0
OUTD=%0000
clear: IF IN5=1 THEN red 'sense reject push button
GOTO clear
check:IF IN6=0 THEN clk 'sense sensor
IF IN2=0 THEN gone 'sense clear push button
GOTO check
clk: IF IN6=1 THEN jump 'pulse off sensor
GOTO clk
jump: cnt=cnt+1 'count number of sensing
IF cnt=4 THEN gone
IF cnt=3 THEN out
IF cnt=2 THEN green
OUTD=%0010 'yellow green: OUTD=%0100 ‘green
GOTO check GOTO check
out: OUTD=%1000 'orange & red: OUTD=%0001 ‘red
GOTO check 'solenoid GOTO check

3.3 Conveyor Inspection Using Shift Registers

The circuit implements shift registers to control a diverter solenoid. Four negative
edge-triggered J-K flip flops as well as a slotted optoisolator, are the main electronics
components for the circuit. The far end of the conveyor is where we located the
diverter. We mount small pieces of block indicator at a fixed distance along the
conveyor system. The block indicator will energize the slotted switch at distance
intervals as the conveyor rolls. The start of the conveyor and the diverter end is three
block indicators away. There is indicator block at the starting point of the conveyor
system where we placed the material. We used a reject switch or a proximity sensor
to detect material put on the conveyor system. A proximity sensor is mounted at the
beginning of the conveyor system where we placed the material. The output of the
first J-K flip-flop will light up if something is on the conveyor system. As the
conveyor rolls, the J-K flip-flops will be activated when the indicator block covers
the slotted optoisolator. It will lit up the second LED at the output of the second J-K
3.3 Conveyor Inspection Using Shift Registers 35

Fig. 3.7 Conveyor system

flip-flop if a material is present. When the material travels to the end of the diverter,
the output of the fourth J-K flip-flop will be energised. At the same time, the diverter
solenoid will be energized to divert the material placed on the conveyor system. The
LEDs indicate which zone the material has arrived in the midst of travelling in the
conveying system. If we put two materials subsequently in each different indicator
block, two LEDs will light up each time the indicator blocks energize the slotted
optical sensor. When nothing is on the conveyor, the far end diverter will not be
activated as the last J-K flip-flop outputs a zero (Figs. 3.7, 3.8).

3.4 Modern Speed Control

We can create a small permanent magnet motor control system by pulse-width

modulation technique. Two 555 timer [16] electronic components are required for
the circuitry. The circuit design is as shown in Fig. 3.9 Pulse-width modulation
control system for a small permanent magnet motor. We supply both the timers
with a +15 V DC source. We connect the first timer on the left as a square wave
oscillating circuit while the second timer on the right connects as a one-shot.
The RC differentiator circuit contains the resistor Rdiff and the capacitor Cdiff. The
first 555 timer of the circuit determines the rise and fall time of the deterministic
output signal. Each time the output from the first timer goes low, the differentiator
delivers a fast negative going spike to trigger the input of the second 555 timer, with
the spike bottoms at 0 V. That initiates another pulse output from the second 555
timer. The period for the second timer input is approximately 25 ms. The operation
of the RC differentiator circuit starts with a low voltage to the negative going side of
the Cdiff. The current flows to charge up capacitor Cdiff and reduces input flow to
the timer. It triggers a negative spike to the input of the second timer. When we
charged the Cdiff to full capacity, the input to the second timer is at 0 V. When the
output of the first timer inverts back from the low level output the capacitor Cdiff
discharges. All current flows from Cdiff to the trigger input of the second timer.
 36

Fig. 3.8 Shift register conveyor control

3 Mechatronics
 3.4 Modern Speed Control

Fig. 3.9 PWM control of a small permanent magnet motor

37
38 3 Mechatronics

Fig. 3.10 RC differentiator

pulse signal

25ms

It causes a positive spike to the trigger input of the timer. The rate of discharge
depends on Cdiff and Rdiff (Figs. 3.9, 3.10).
The RC differentiator circuit time constant is 0.15 ms as we used a 1.5 kΩ and
the 0.1 uF passive components. We determined the period by using resistors and
capacitor of 18 kΩ and 1 uF respectively. By the connection of the 0.01 uF
capacitor to the control input, the 555 timer oscillates at 50 % duty cycle, with a
period of 25 ms. The calculation for the timing of the first timer circuit is as follow.

T ¼ 2  0:7  18 k  1u
ð3:1Þ
¼ 25:2 ms:

We connected the resistor and capacitor from the second timer as 10 kΩ and 2.2 uF
respectively.

T ¼ RC ¼ 10 k  2:2 u ¼ 22 ms: ð3:2Þ

As can be seen, it forms the variable time peak of 22 ms. Therefore, the maximum
duty cycle of the second timer is reachable at (22 ms/25 ms) 88 % efficiency. The
40 v permanent magnet motor is controllable with a maximum duty cycle of 88 %.
The motor’s duty cycle is variable from 10 to 88 % by varying the control source to
the second one-shot timer from 1.8 to 10 V. We control the mechanical shaft power
of the permanent magnet motor by the control source. A +40 V DC source powered
the permanent magnet motor. Low pass filter can be installed to the motor when
needed.
Chapter 4
Microcontroller

Microcontrollers require the digital signal to perform functions and applications.

For ensuring the analog input signals matched that of the digital input of the
microcontroller, the analog signal must be within the range of the digital device
with TTL or CMOS logic level compatibility. Moreover, the output impedance of
the sending device should be small enough for the transfer of power to the receiving
device. Conversely, the input device should have a high input impedance to match
the output device. It is related to the driving current specifications for the output of
the sending devices and the sinking current of the input receivers.

4.1 Basic I/O Modules

Load and run program 4.1. Note that port A connector is internally being connected
to 10 kΩ pull-up resistors. Observe how to read in a 8 bits information via Port A.
Note that the address of Port A is $1000. DDRA address at $1001. The register
controls whether the bit 0–7 use as input (Low) or output (High) pins. Index
addressing with register X. CLI (clear interrupt mask) to enable interrupts by setting
I bit to zero. By doing so, we enable the software debugger. Check the value at
address $0050 using the debugger. It should give you the value indicated by the dip
switches. All files associate with the file extension .a11 in the assembly format
(Fig. 4.1).

© Springer Science+Business Media Singapore 2016 39

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_4
40 4 Microcontroller

Fig. 4.1 8-bit dip switch input circuit

---------------------------------------
Program 4.1: Read Dip Switch
---------------------------------------
porta equ 0
ddra equ 1
org $8000 ; Micro-P starting address
cli ; enable the debugger
ldx #$1000
ldaa #$00
staa ddra, x ;set poara as input
again ldaa porta, x ;load dip switch status
staa $0050 ;store dip switch status
bra again
----------------------------------------------
Program 4.2: Write to Output LEDs
----------------------------------------------

portb equ $60 staa portb, x

org $8000 jsr delay
cli bra again
ldx #$1000 org $8200
again ldaa #$0f delay ldy #$ffff
staa portb, x loop dey #$1
jsr delay bne loop
ldaa #$f0 brts
4.1 Basic I/O Modules 41

Fig. 4.2 LEDs output circuit

Note that port B (Fig. 4.2) of the output circuit is connected internally to
ULN2803A Darlington transistors. You can load and run the program 4.2 to
observe the following points. How to output a 8 bits output via Port B. The address
of Port B is $1060. That is it is Port B of the basic microprocessor, not the Port B of
the 68HC11 microcontroller. We observe the branching and looping techniques.

----------------------------------------------
Program 4.3: Push Button & Buzzer
----------------------------------------------

portd equ 8
ddrd equ 9

org $8000 ; Micro-P starting address

cli ; enable the debugger
ldx #$1000
ldaa #%00100000 ;set pd2 as input & pd5 as output
staa ddrd, x
begin bclr portd, x $20 ;clear bit pd5
again brset portd, x $04 begin ;check bit pd2
bset portd, x $20 ;set bit pd5
bra again

We connect connector D to port D and port G (Fig. 4.3). The function of

program 4.3 is to activate a buzzer using a push button. Make sure you understand
the following points. Address of port D is $1008. Address of Data Direction
Register for port D (DDRD) is $1009. This register controls whether bit 0–5 use as
input (LOW) or output (HIGH) pins. Bit 6 and 7 are not usable. Index addressing
42 4 Microcontroller

Fig. 4.3 Buzzer and push button circuit

with register X. We used branching and looping. The high current Darlington
transistor arrays are for driving the buzzer load.
In basic, both programs 4.4 and 4.5 are the same. They read in the PE2 port for
the A/D conversion from the potentiometer connected to it (see Fig. 4.4). Program
4.6 can read in any selected port to be read (PE0 to PE3). In the case, PE2 is being
read at the address $1033 in the program. In program 4.7, we measure at pin6 of
portA the sampling period.

Fig. 4.4 A/D converter

4.1 Basic I/O Modules 43

---------------------------------------
Program 4.4: A/D Converter 1
---------------------------------------
org $0100
adctl equ $30 ; option registerr
addr equ $33 ; store conversion result
begin: ldx #$1000 ; reads in analog 0 – 5v into PE2 channel
bset $39, x $80
bclr $39, x $40
off: ldab #$02
stab adctl, x ; select channel PE2 clr scan
check: ldaa adctl, x ; and mult bits of adctl
brclr adctl, x $80 check ; read status until conversion ends
ldaa addr, x ; If CCF=1 read results in ADR3 register
staa $ec00
bra off ; halt here
org $fffe
fdb begin

---------------------------------------
Program 4.5: A/D Converter 2
---------------------------------------
org $0100
begin: ldx #$1000 ; reads in analog 0 – 5v into PE2 channel
bset $39, x $80
bclr $39, x $40
off: ldab #$02
stab $1030 ; select channel PE2 clr scan
check: ldaa $1030 ; and mult bits of adctl
bpl check ; read status until conversion ends
ldaa $1033 ; If CCF=1 read results in ADR3 register
staa $ec00
bra off ; halt here
org $fffe
fdb off
44 4 Microcontroller

---------------------------------------
Program 4.6: A/D Converter 3
---------------------------------------
org $0100
adctl equ $30 ; option registerr
addr equ $33 ; store conversion result
begin: ldx #$1000 ; reads in analog 0 – 5v into PE2 channel
bset $39, x $80 ; bit in ADPU option reg to
bclr $39, x $40 ; power up and use E clk
off: ldab #$30 ; use PE0-PE3 mult continuous scan=1
stab adctl, x ; select channel PE2 clr scan
again: ldaa adctl, x ; and mult bits of adctl
brclr adctl, x $80 again ; read status until conversion ends
ldaa addr, x ; channel PE2 at $1033
staa $ec00
bra again ; get next continuous scan
org $fffe
fdb begin

---------------------------------------
Program 4.7: A/D Converter 4
---------------------------------------
org $0100
adctl equ $30 ; This program measures
addr equ $33 ; the sampling period by
begin: ldx #$1000 ; sending 1 out pin A6
ldaa #$40 ; pin 6
bset $39, x $80 ; bit in ADPU option reg to
bclr $39, x $40 ; power up and use E clk
off: ldab #$30 ; use PE0-PE3 mult continuous scan=1
stab adctl, x ; select channel PE0 – PE3 clr scan
again: ldaa adctl, x ; and mult bits of adctl
brclr adctl, x $80 again ; read status until conversion ends
staa $1000 ; pulse A6
eora #$40 ; reverse pulse A6
undone:bra again ; get next continuous scan
org $fffe
fdb begin

4.2 LCD and Keypad

Next, the LCD program 4.8 teaches how to program in assembly language the
following extension .a11 assembly file. How to display a character of 8 bits word on
LCD, via port M, 4 bits nibble at a time. You can replace “Welcome to the Lab”
with own text. You also learn how to move the cursor to the start of each line. The
flow chart illustrates the flow of steps involved in the program. The LCD used is a
16 × 2 characters module (Flowchart 4.1).
4.2 LCD and Keypad 45

Flowchart 4.1 LCD display

46 4 Microcontroller

----------------------------------
Program 4.8: LCD Display
----------------------------------

PORTB equ $1060

PORTC equ $1061
PORTM equ $1062
PORTN equ $1063
sect data
org $2000
sect text
org $8000
lds #$7FFF
cli ; enable debugger
bsr lcdinit ; initialize LCD
ldx #txt1 ; starting address of line 1 message --› index X
ldy #lcdbuf ; starting address of buffer --› index Y
main1:
ldab 0, x
beq main2 ; search for “0” – end of message
stab 0, y ; transfer line 1 into first 16 bytes of buffer
inx
iny
bra main1
main2:
ldx #txt2 ; starting address of line 2 message --› index X
ldy #lcdbuf+16 ; 16 + starting address of buffer --› index Y
main3:
ldab 0, x
beq main4 ; search for “0” – end of message
stab 0, y ; transfer line 2 into next 16 bytes of buffer
inx
iny
bra main3
main4:
bsr lcdput ; go subroutine for LCD display
end: bra end
; message for display
txt1: fcb ‘Welcome’, 0 ; line 1 message
txt2: fcb ‘to the Lab.’, 0 ; line 2 message
4.2 LCD and Keypad 47

; ****************Subroutine************************
; lcdnib, lcdnibw – send code to LCD through PORTM
; lcdtempo – delay
; lcdputc – arrange command code to be sent
; lcdinit – set up LCD
; lcdput – arrange character code to be sent
sect data
lcdbuf: rmb 32 ; reserve 32 bytes of ram as buffer for storing
; characters to be displayed
sect text
lcdnib: ; send code to LCD through PORTM (pm0 – pm3)
andb #$4F
orab #$30
stab PORTM
tba
oraa #$80
staa PORTM ; 1 to 500ns
stab PORTM
rts
lcdnibw:
bsr lcdnib
ldd #20 ; delay 40us (20 x 2us)
bsr lcdtempo
rts
lcdtempo:
subd #1 ; 1unit in reg D x 2us (12MHz)
bne lcdtempo
rts
lcdputc: ; arrange command code – send 4 bits at one time
; to LCD screen, higher nibble goes first
pshb ; retain the content of ACCB
lsrb ; shift out lower nibble
lsrb
lsrb
lsrb
bsr lcdnib; send higher nibble
pulb ; restore ACCB
bsr lcdnibw ; send lower nibble
rts
lcdinit: ; set up the LCD for 4 lines interface, 2 line mode,
; 5x7 dot format, increment, display shift off.
Ldab #$3F
stab PORTM
ldd #8000
bsr lcdtempo ; 15ms
ldab #3
48 4 Microcontroller

bsr lcdnib ; LCD ‹-- 3

ldd #2000
bsr lcdtempo ; 4ms
ldab #3
bsr lcdnib ; LCD ‹-- 3
ldd #50
bsr lcdtempo ; 0.1ms
ldab #3
bsr lcdnibw ; LCD ‹-- 3, 40us
ldab #2
bsr lcdnibw ; LCD ‹-- 2, 40us
ldab #$29
bsr lcdputc ; LCD ‹-- 29
ldab #$08
bsr lcdputc ; LCD ‹-- 08, display off
ldab #$01
bsr lcdputc ; LCD ‹-- 01, clear
ldd #2000
bsr lcdtempo
ldab #$06
bsr lcdputc ; LCD ‹-- 06
ldab #$0E
bsr lcdputc ; LCD ‹-- 0E, display on, cursor
ldx #lcdbuf ; clear the buffer
ldab #‘ ‘
lcdin1: stab 0, x
inx
cpx #lcdbuf+32
bne lcdin1
rts
lcdput: ; transfer content of buffer to LCD
ldab #$80 ; move cursor to the beginning of first line
bsr lcdputc
ldx #lcdbuf
lcdp1: ; cursor at end of line 1?
Cpx #lcdbuf+16
bne lcdp2
ldab #$A8 ; if yes, move cursor to the start of 2nd line
bsr lcdputc
lcdp2: ldab 0, x ; arrange and send the character code – 4 bits at one
; time, higher nibble goes first
4.2 LCD and Keypad 49

lsrb
lsrb
lsrb
lsrb
orab #$40
bsr lcdnib ; send higher nibble
ldab 0, x
orab #$40
bsr lcdnibw ; send lower nibble
inx
cpx #lcdbuf+12
bne lcdp1
rts

The next program for the keypad and LCD demonstrates how to read in char-
acters from Port D & G of connector D, 4 bits at a time. You will learn how to
change the content of the conversion table and check the LCD character map. The
keypad used is a 16 keys keypad. We use a keypad encoder as shown in Fig. 4.5,
for the keypad connections (Flowchart 4.2).
 50
4 Microcontroller

Fig. 4.5 Keypad and LCD using keypad encoder

4.2 LCD and Keypad 51

Flowchart 4.2 Keypad and LCD

52 4 Microcontroller

-------------------------------------
Program 4.9: Keypad & LCD
-------------------------------------

PORTD equ $1008

DDRD equ $1009
PORTG equ $1002
DDRG equ $1003
PORTM equ $1062
; initialize the program counter, stack pointer, LCD,
; Port D and Port G direction registers.
sect text
org $8000 ; program starts at $8000
lds #$7FFF
cli ; enable debugger
bsr lcdinit ; initialize LCD
ldab #$00
stab DDRD ; set portd as inputs
ldab #$03
stab DDRG ; set portg bit 2–7 as inputs
; start checking keypad entry and display characters
start: ldy #0 ; use index Y to keep track the cursor position
ldab #$01 ; clear display & home cursor: command code “$01”
bsr lcdputc
ldd #2000 ; delay
bsr lcdtempo
loop: cpy #32 ; checking cursor position – is it at the end of line 2?
beq start ; if “Yes”, clear and home
cpy #16 ; checking cursor position – is it at the end of line 1?
beq loop1 ; if “Yes”, move cursor to the beginning of line 2 –
; command code “$A8”
bra next
loop1: ldab #$A8
bsr lcdputc
next: ldaa PORTG ; check data available signal at pg3
staa $50
brclr $50 $08 next ; “high” to ensure valid keypad entry
ldab PORTD ; read portd
andb #%00111100 ; keep the content of pd2 – pd5 and mask out
; the rest
lsrb ; shift pd2 -- pd5 to lower nibble
lsrb
ldx #table1 ; starting address of conversion table --›index X
abx ; get the actual character to be displayed, (X) +
; (ACCB) --› (X)
4.2 LCD and Keypad 53

ldab 0, x ; load character code from conversion table and

bsr lcdput_keypad ; branch to subroutine for LCD display
ldd #$ffff ; delay – prevent multiple entry
bsr lcdtempo
ldd #$ffff ; delay again
bsr lcdtempo
iny
bra loop
; set the conversion table
table1: fcb $31, $32, $33, $46, $34, $35, $36, $45, $37, $38, $39, $44, $41,
$30, $42, $43
;***************Subroutine*****************************
; lcdnib, lcdnibw – send code to LCD through PORTM
; lcdtempo – delay
; lcdputc – arrange command code to be sent
; lcdinit – set up LCD
; lcdput_keypad – arrange character code to be sent
lcdnib: ; send code to LCD through PORTM (pm0 – pm3)
andb #$4F
orab #$30
stab PORTM
tba
oraa #$80
staa PORTM
stab PORTM
rts
lcdnibw:
bsr lcdnib
ldd #20
bsr lcdtempo
rts
lcdtempo: ; delay
subd #1
bne lcdtempo
rts
lcdputc: ; arrange command code – send 4 bits at one time,
; higher nibble first
pshb ; retain the content of ACCB
lsrb
lsrb
lsrb
lsrb
bsr lcdnib ; send higher nibble
pulb ; restore ACCB
bsr lcdnibw ; send the lower nibble
rts
54 4 Microcontroller

lcdinit: ; set up the LCD for 4 lines interface, 2 line mode,

; 5x7 dot format, increment, display shift off
ldab #$3F
stab PORTM
ldd #8000
bsr lcdtempo
ldab #3
bsr lcdnib
ldd #2000
bsr lcdtempo
ldab #3
bsr lcdnib
ldd #50
bsr lcdtempo
ldab #3
bsr lcdnibw
ldab #2
bsr lcdnibw
ldab #$29
bsr lcdputc
ldab #$08
bsr lcdputc
ldab #$01
bsr lcdputc
ldd #2000
bsr lcdtempo
ldab #$06
bsr lcdputc
ldab #$0E
bsr lcdputc
rts
lcdput_keypad: ; arrange character code – send 4 bits at a time, higher
; nibble first
stab $0051 ; keep content of ACCB in location $0051
lsrb ; shift out lower nibble
lsrb
lsrb
lsrb
orab #$40
bsr lcdnib ; send higher nibble
ldab $0051; restore ACCB
orab #$40
bsr lcdnibw ; send lower nibble
rts

The following program (4.10) in C language performs the function of calculating

the square root of a number we key in. You may read the input from the keypad,
and the answer for the square root will display on the LCD screen. As shown in the
diagram, the 16-key keypad input goes into the portN and portM of the micro-
controller connector. The connector LCD links to portM and portN of the micro-
controller. The board trimmer P3 controls the brightness of the LCD. The answer in
float number will display via portM, 4 bit (nibble) at a time. Compare the schematic
 4.2 LCD and Keypad

Fig. 4.6 Keypad and LCD

55
56 4 Microcontroller

diagram with the other to find out the complexities of the other keypad and LCD
circuit diagram. You may also note the simplicity of using the high levels language
instead of the assembly language (Fig. 4.6).

--------------------------------------------------------
Program 4.10: Keypad & LCD In C Program
--------------------------------------------------------
#include “hc11.h”
#define PORTM *(unsigned char *)(_IO_BASE + 0x62)
#define PORTN *(unsigned char *)(_IO_BASE + 0x63)
#include <stdio.h>
#include <math.h>
void lcdinit(void);
char keyget(void);
double e;

void main(void)
{ int i;
lcdinit( );
for (i=2;;)
{ e = sqrt(i);
if (i>=0)
print(“ sqrt(%d)=%f”, i, e);
i= keyget( ) – ‘0’; } }

/**************************Lcd*****************************
//PORTM bit7=LCD EN = 0 by default
//PORTM bit4=DAC CS = 1 by default
#define LCDCOLS 16
#define LCDLINES 2
#define LCDBUFSZ (LCDCOLS*LCDLINES)
char lcdbuf[LCDBUFSZ];

void lcdwait(int cnt)

{ int i;
for (; cnt; cnt--)
for (i=0; i<1000; i++); }
void lcdnib( unsigned char c)
{ c = (c & 0x4F) | 0x30;
PORTM = c | 0x80; /*1 x 500ns*/
PORTM = c; }
void lcdnibw(unsigned char c)
{ lcdnib (c);
lcdwait(0); }
void lcdputc(unsigned char c)
{ lcdnib((c>>4)&0xF);
lcdnibw(c&0xF); }
void lcdinit(void)
{ char *s;
PORTM = 0x3F;
lcdwait(5); /* 15ms */
4.2 LCD and Keypad 57

lcdnib(3); /* lcd = 3 */
lcdwait(1); /* 4.1ms */
lcdnib(3); /* lcd = 3 */
lcdwait(0); /* 0.1 ms */
lcdnibw(3); /* lcd = 3 x 40us */
lcdnibw(2); /* lcd = 2 x 40us */
lcdputc(0x28 | (LCDLINES-1)); /*LCDLINES:info for the sim11*/
lcdputc(0x08); /* display off */
lcdputc(0x01); /* clear display */
lcdwait(1);
lcdputc(0x06);
lcdputc(0x0E); /* display on, cursor */
for (s=lcdbuf; s<lcdbuf + LCDBUFSZ; s++)
*s = ‘ ‘;
putchar(‘ ‘); }
int putchar(char c) /*library function*/
{ char *s;
for (s=lcdbuf; s<lcdbuf + LCDBUFSZ – 1; s++)
*s = s[1];
*s = c;
lcdputc(0x80);
for (s = lcdbuf; s<lcdbuf + LCDBUFSZ; s++)
{ if (s==lcdbuf + (LCDBUFSZ/2))
lcdputc(0x80 + 0x28);
lcdnib( (*s >>4) | 0x40);
lcdnibw(*s | 0x40); }
return c; }

/**************Keyboard*************************/
char keyget(void)
{ static char old_key;
char x = 0;
#if 0 // keypad 3 x4
PORTM = 0x37; /* K=0 J=1 H=1 G=1 */
if ((PORTN&0x08)==0) x = ‘1’;
if ((PORTN&0x04)==0) x = ‘2’;
if ((PORTN&0x02)==0) x = ‘3’;
PORTM = 0x3B; /* K=1 J=0 H=1 G=1 */
if ((PORTN&0x08)==0) x = ‘4’;
if ((PORTN&0x04)==0) x = ‘5’;
if ((PORTN&0x02)==0) x = ‘6’;
PORTM = 0x3D; /* K=1 J=1 H=0 G=1 */
if ((PORTN&0x08)==0) x = ‘7’;
if ((PORTN&0x04)==0) x = ‘8’;
if ((PORTN&0x02)==0) x = ‘9’;
PORTM = 0x3E; /* K=1 J=1 H=1 G=0 */
58 4 Microcontroller

if ((PORTN&0x08)==0) x = ‘*’;
if ((PORTN&0x04)==0) x = ‘0’;
if ((PORTN&0x02)==0) x = ‘#’;
#else // keypad 4 x 4
PORTM = 0x37; /* K=0 J=1 H=1 G=1 */
if ((PORTN&0x08)==0) x = ‘C’;
if ((PORTN&0x04)==0) x = ‘B’;
if ((PORTN&0x02)==0) x = ‘0’;
if ((PORTN&0x01)==0) x = ‘A’;
PORTM = 0x3B; /* K=1 J=0 H=1 G=1 */
if ((PORTN&0x08)==0) x = ‘D’;
if ((PORTN&0x04)==0) x = ‘9’;
if ((PORTN&0x02)==0) x = ‘8’;
if ((PORTN&0x01)==0) x = ‘7’;
PORTM = 0x3D; /* K=1 J=1 H=0 G=1 */
if ((PORTN&0x08)==0) x = ‘E’;
if ((PORTN&0x04)==0) x = ‘6’;
if ((PORTN&0x02)==0) x = ‘5’;
if ((PORTN&0x01)==0) x = ‘4’;
PORTM = 0x3E; /* K=1 J=1 H=1 G=0 */
if ((PORTN&0x08)==0) x = ‘F’;
if ((PORTN&0x04)==0) x = ‘3’;
if ((PORTN&0x02)==0) x = ‘2’;
if ((PORTN&0x01)==0) x = ‘1’;
#endif
if (x == old_key)
return 0;
old_key = x; return x; }

4.3 Waveform Timings

We can appreciate the use of the microcontroller (68HC11) timing system to

generate a square waveform. The program 4.11 shows the square wave generator.
The square wave generated from portA bit 5 (PA5) can display on the oscilloscope.
We alter the period of the waveform by changing the value ‘#4000’ in the interrupt
subroutine.
4.3 Waveform Timings 59

-------------------------------------------
Program 4.11: Square Wave Timer
-------------------------------------------
regbas equ $1000 ; bas address
porta equ $0 ; port A
tcnt equ $0e ; timer counter
toc3 equ $1a ; timer output compare 3
tmsk1 equ $22 ; timer interrupt mask 1
tflg1 equ $23 ; timer interrupt flag 1
tctl1 equ $20 ; timer control 1
; outputsquare wave signal at bit5 of porta
org $8000 ; program starts at $8000
cli
main: ldx #regbas
ldaa #%00100000
staa tmsk1, x ; OC3I interrupt enable
staa tflg1, x ; send a one to clear the OC3F flag
ldaa #%00010000
staa tctl1, x ; OM3:OL3=0:1;toggle TOC3 output after a successful compare
ldd tcnt, x ; load current content of timer counter
addd #2000 ; added by 2000 counts (333nsec per count, total up 0.67msec)
std toc3, x ; store the result in compare register TOC3
undone: bra undone ; do something useful here
;***********interrupt subroutine*******************************
rtoc: bset tflg1, x $20 ; send a “one” to clear the OC3F flag
ldd #4000 ; set high or low period for 400x333 = 1.332msec
addd tcnt, x ; store the value in compare register
std toc3, x ; interrupr generated when next tcnt = toc3
rti
;**************define interrupt vector address********************
org $ffe4
fdb rtoc

We can program a pulse width modulation signal to drive a DC motor. The

output for portA bit5 of the microcontroller connects to the motor. Observe that we
adjust the speed of the motor by varying “hi_time” between 1000 and 7000. At the
label low in the interrupt subroutine, the value #7500-hi_time is added to
Accumulator D.
60 4 Microcontroller

-------------------------------------------------------------------
Program 4.12: PWM Signal of 400Hz at 2.5ms Period
-------------------------------------------------------------------
regbas equ $1000 ; bas address
porta equ $0 ; port A
tcnt equ $0e ; timer counter
toc3 equ $1a ; timer output compare 3
tmsk1 equ $22 ; timer interrupt mask 1
tflg1 equ $23 ; timer interrupt flag 1
tctl1 equ $20 ; timer control 1
hi_time equ 3000 ; high period 3000 cnts (333ns per count)
org $8000 ; program starts at $8000
cli
main: bsr initoc3 ; set up timer output compare 3, TOC3
undone: nop
bra undone

;*******************subroutine**********************
Initoc3: ldx #regbase
ldaa #$20
staa porta, x ; set PA5 high first
staa tmsk1, x ; OC3I interrupt enable
staa tflg1, x ; send a “one” to clear OC3F interrupt flag
staa tctl1, x ; OM3: OL3 = 1:0; set PA5 low upon interrupt
ldd tcnt, x ; load current content of timer counter
addd #50 ; added by 50 counts, initial setup before interrupt occurs
std toc3, x ; store the result in compare register TOC3
rts

;***********interrupt subroutine*******************************
rtoc3: ldx #$1000
ldaa porta, x ; check porta PA5 high or low
bita #$20
beq low
high: ldd tcnt, x
4.3 Waveform Timings 61

addd #hi_time; set the new value to be compared = high period

std toc3, x ; store the value in TOC3 register
ldaa #$20
staa tcntl1, x ; OM3 : OL3 =1:0; set PA5 low upon interrupt
bra undone
low: ldd tcnt, x ; period of 2.5ms (7500 x 333ns)
addd #7500-hi_time ; set the new value to be compared = low period
std toc3, x ; store the value in TOC3 register
ldaa #$30 ;
staa tcntl1, x ; OM3 : OL3 =1:1; set PA5 high upon next OC3 interrupt
done: ldaa #$20 ; clear OC3F interrupt flag
staa tflg1, x
rti

;**************define interrupt vector address********************

org $ffe4
fdb rtoc3

The next program can perform three functions with the external connections.
Refer to the circuit diagram for the program. The optical or slotted switch (con-
nected to PA0) is used as an input device to start the motor. After starting the motor,
the microcontroller can ignore the input signal from the slotted switch. Note 1: It is
important to note that to establish the correct initial status of the slotted switch the
power supply to the optical switch and motor must be turned on before turning on
the microcontroller. Note 2: After you have loaded the program, sometimes
pressing ‘Go’ does not work as expected. If this happens, press the Reset switch
instead of ‘Go’.
The external rotary potentiometer connects to the A/D converter (PE2) port E.
We use the potentiometer to control the motor speed. Depending on the converted 8
bit reading at PE2, three distinct speed levels are set accordingly.

PE2 reading (Hex) Value of “h_time”

00–40 1000
41–C0 4000
C1–FF 7000

Note: Under the label ‘low’ in the interrupt subroutine, the value
(#7500-hi_time) should be added to Accumulator D. You may refer to the diagram
in Fig. 4.7 for the PWM control of DC motor. The A/D pulse width varying circuit
is the same as Fig. 4.4 A/D converter. Port E is only a 8 bit read in port for A/D
converter. The third function is the inclusion of the XIRQ interrupt subroutine to
stop the motor. We connect a push button to the X connector’s XIRQ pin of the
microcontroller unit.
62 4 Microcontroller

start motor

stop motor

Fig. 4.7 PWM control of DC motor

4.3 Waveform Timings 63

------------------------------------------------------------------------
Program 4.13: PWM Motor Control With A Potentiometer
------------------------------------------------------------------------------------

; pwm output at porta bit5 (PA5)

; pulse period of 2.5ms (7500 cnts x 333ns), pulse frequency is 400Hz
regbas equ $1000 ; base address
porta equ $0 ; port A
ddra equ $1
tcnt equ $0e ; timer counter
toc3 equ $1a ; timer interrupt mask 1
tflg1 equ $23 ; timer control 1
adct1 equ $30 ; ADCTL register
adr4 equ $34 ; A?D result register
optn equ $39 ; option register
portg equ $02
ddrg equ $03
hi_time equ 3000 ; high period, 3000 cnts (333ns per count)
org $8000 ; program starts at $8000
clra
tap ; transfer ACCA to CCR – to clear interrupt M
cli
begin: ldx #regbase
ldaa #$FF
staa ddrg, x
ldab #$FF
stab portg, x ; turn off red and green LEDs
bset optn,x $80; enable A/D conversion system – set bit 7
bsr delay ; delay of 100 ms is required to stabilize A/D converter
bclr optn, x $40 ; use sys clock E
ldab #$22 ; perform continuous scan & single channel conversion (PE2
only)
stab adctl,x ; at the same time, start the conversion by writing to ASCTL reg.
ldaa #$20
staa tctl1, x
swtich1: brclr porta, x $01 switch1 ; check opto switch
switch2: brset porta, x $01 switch2 ; off-on-off sequence
main: bsr initoc3 ; set up timer output compare 3, TOC3
undone: nop ; do something interesting here
bra undone
;********************subroutine*******************
Initoc3: ldx #regbase
ldaa #$20
staa porta, x ; set PA5 high first
staa tmsk1, x ; OC3I interrupt enable
staa tflg1, x ; send a “one” to clear OC3F interrupt flag
staa tctl1, x ; OM3:OL3=1:0; set PA5 low upon interrupt
64 4 Microcontroller

ldd tcnt, x ; load current content of timer counter

addd #50 ; added by 50 counts, initial setup before interrupt occurs
std toc3, x ; stored result in TOC3 compare reg.
rts
delay: ldaa #240
loop: suba #1
bne loop
rts
;************interrupt subroutine*******************
rtoc3: ldx #regbas
again: brclr adctl, x $80 again ; check status and wait for end of conversion – bit7
ldaa adr4, x ; transfer the converted value from A/D result register ADR4
cmpa #$C0 ; check PE2 >= 3.75v
bhs sp3
cmpa #$80 ; check PE2 >=2.5v
bhs sp2
cmpa #$40 ; check PE2 >= 1.25v
bhs sp1
sp1: ldd #1000 ; speed 1 ‘ON’ width
std $0050
ldd #6500 ; speed 1 ‘OFF’ width
std $0052
bset portg, x $03
bra next
sp2: ldd #4000 ; speed 2 ‘ON’ width
std $0050
ldd #3500 ; speed 2 ‘OFF’ width
std $0052
bset portg, x $03
bclr portg, x $01
bra next
sp3: ldd #7000 ; speed 3 ‘ON’ width
std $0050
ldd #500 ; speed 3 ‘OFF’ width
std $0052
bclr portg, x $03
next: ldaa porta, x ; check PA5 high or low
bita #$20
beq low
high: ldd tcnt, x
addd $0050 ; set the new value to be compared = high period
std toc3, x ; store the value in TOC3 register
ldaa #$20
staa tctl1, x ; OM3:OL3=1:0; set PA5 low upon interrupt
bra done
low: ldd tcnt, x
addd $0052 ; set the new value to be compared = low period
4.3 Waveform Timings 65

std toc3, x ; store the value in TOC3 register

ldaa #$30
staa tctl1, x ; OM3:OL3=1:1;next OC3 interrupt set pin PA5 high
done: ldaa #$20 ; clear OC3F interrupt flag
staa tflg1, x
rti
;***********interrupt subroutine (XIRQ)*******************
Xirq_sub: ; do something here
Ldx #regbas
Ldaa #$20 ; PA5 low to stop motor
staa tctl1, x
bclr portg, x $03
check1: brclr porta, x $01 check1 ; wait for opto to sense then start motor again
check2: brset porta, x $01 check2 ; off-on-off sequence
bset portg, x $03
rti
;*******define interrupt vector address****************
org $ffe4
fdb rtoc3
org $fff4 ; XIRQ (active low)
fdb xirq_sub

4.4 Pressure Sensing

Smart sensors can adjust their interface outputs to the receiving appliances. But
there are factors exist that produce errors in the measuring sensors. Factors to
consider are scaling (static range) and offset voltages, drift due to temperature after
prolonged operations in a non-cooled space, noise margins and finally, bandwidth
for dynamics, frequency-related measurements. Therefore, signal conditioning
circuits are often necessary for better interpretation and presentation of the
incoming signal from the sensors into the microcontroller.
The setup in the figure measures the air pressure in the host line using the
SCX01DNC pressure transducer. Verify the resistors are correct to give the desired
output equations. We can vary the input pressure and plot the digital values
 66

Fig. 4.8 Pressure sensing circuit

4 Microcontroller
4.4 Pressure Sensing 67

converted by the 68HC11 A/D converter. The oscilloscope can be used to find the
time delay from the instance of pressure change to the time of complete digital
conversion by the processor, that is, the duty cycle of the system. Most A/D
converters lose their accuracy when the voltages come close to the rail values. We
can improve it by changing the resolution to the A/D converter. The circuit for the
pressure detector requires a gain to amplify the signal large enough to view or use
for other activities. The amplifiers simply do the job. Any of the A/D converter
programs (P4.4–P4.7) can be used to measure the converted signal values
(Fig. 4.8).

4.5 Temperature Measurement

Why is the standard 741 Op Amp not good enough? It gives an offset of 6 mV
and drift of 15 mV/°C. The BJT is also susceptible to noise from RF rectification.
So, let say we want to amplify the signal to a magnitude of 100, for a temperature
sensing of 100 mV/°C. We will have an offset of 6 °C and a drift of ±7 °C! We
must use a more precise CMOS-based amplifier. The circuit for the temperature
measurement is as shown in the diagram. AD590 is the temperature sensor. The
resistors are chosen to give a slope of 5. Trim pots are necessary to ensure this as
well as to remove offset voltage at 0 °C. Use a hot blower to increase the
measured temperature. We measure the Vo into the portE of the microcontroller.
Vo = (50 mV/°C) * T, where T represents the temperature measured. We can
make use of any of the A/D converter program 4.4–4.7 to detect the signal into
the microcontroller unit. We can then compare and verify the voltage signal with
a thermometer (Fig. 4.9).

4.6 Stepper Motor Control

We can program the 68HC11 microcontroller unit to drive a stepper motor. The
stepper motor we use needs a 5 V input power supply. Power MOSFETs for
example, the BUZ12 n-channel transistors are used to switch on the stepper
motor. The terminals of Port A of the MCU connects to output to the gates of
each transistor. The stepper motor has four coils to be activated in a sequential
 68

Fig. 4.9 Temperature sensing circuit

4 Microcontroller
4.6 Stepper Motor Control 69

manner to rotate the motor. IN4001 diodes are used as free-wheeling diodes to
protect the motor coils when the transistors are de-activated. It helps to relieve
the stored charges in the coils when the transistors are turned off. So the motor
coils are protected from burnings. Three methods used can activate and energise
the motor. They are the full phase, full step and half step modes. The three
programs written in assembly language for the 68HC11 microcontroller illus-
trates the different techniques used to energise the four motor coils for rotating
the motor. Only the correct sequence of turning on the phase of the coils will
activate the motor to turn in a continuous direction. The 68HC12 family has
slightly different configurations in the registers. We can just change the data
direction register, and the data port register addresses in the program, to perform
the same function (Fig. 4.10).

Fig. 4.10 Stepper motor driving circuit

70 4 Microcontroller

------------------------------------------
Program 4.14: Two Phase Mode
------------------------------------------

;Program to drive a 200step/rev unipolar stepper motor in two-phase on

;mode(200steps/rev).
;Two-phase on mode gives higher torque at the expense of higher current.
;two-phase on mode(cw) = coilsA+B then coilsB+C then coilsC+D then
;coilsD+A then repeat cycle.
;two-phase on mode(ccw)= coilsC+D then coilsB+C then coilsA+B then
;coilsD+A then repeat cycle.
;The motor shaft rotates clockwise by 1rev(200steps) then ;counterclockwise by
;1rev
;A short delay separates each cw and ccw motion.
;This is repeated for ever.
;The stepping rate (steps/sec) is set by the delay subroutine
;coilA connected to pa7
;coilB connected to pa6
;coilC connected to pa5
;coilD connected to pa4
;start initialisation
org $e000
cli
; set bit7 of PACTL to output
ldaa #$80
staa $1026 ;bit7=1 in DDRA7
;finish initialisation
;rotate shaft cw by 1 rev (200 steps)
;use the Y index reg as a counter
again ldy #50 ;200steps=50 x 4 where 50 is loaded into IY and
;4=four steps of cycle
revcw jsr cw
dey
bne revcw

;now give short delay before reversing (share the use of the delay subroutine)
ldy #25
shdel1 jsr delay
dey
bne shdel1
;now rotate shaft ccw by 1 rev (200 steps)
;again use the Y index reg as a counter
ldy #50
revccw jsr ccw
dey
bne revccw
4.6 Stepper Motor Control 71

;now give another short delay before reversing again(share the use of the delay
;subroutine)
ldy #25
shdel2 jsr delay
dey
bne shdel2
jmp again
;subroutine to step through 1 cycle of full-step mode in
;clockwise direction
cw ldaa #$c0 ;step1 of cycle(coilsA+B=on)
staa $1000
jsr delay
ldaa #$60 ;step2 of cycle(coilsB+C=on)
staa $1000
jsr delay
ldaa #$30 ;step3 of cycle(coilsC+D=on)
staa $1000
jsr delay
ldaa #$90 ;step4 of cycle(coilsD+A=on)
staa $1000
jsr delay
rts
;subroutine to step through 1 cycle of full-step mode in
;counter clockwise direction
ccw ldaa #$30 ;step3 of cycle(coilsC+D=on)
staa $1000
jsr delay
ldaa #$60 ;step2 of cycle(coilsB+C=on)
staa $1000
jsr delay
ldaa #$c0 ;step1 of cycle(coilsA+B=on)
staa $1000
jsr delay
ldaa #$90 ;step4 of cycle(coilsD+A=on)
staa $1000
jsr delay
rts
;delay routine works close to pull-in speed of motor
delay ldx #500 ;
del1 dex
bne del1
rts
72 4 Microcontroller

------------------------------------
Program 4.15: Full Step Mode
------------------------------------

;Program to drive a 200step/rev unipolar stepper motor in full-step mode.

;Full-step mode(cw) = coilA then coilB then coilC then coilD then repeat cycle.
;Full-step mode(ccw)= coilD then coilC then coilB then coilA then repeat cycle
;The motor shaft rotates clockwise by 1rev(200steps) then counterclockwise by
;1rev
;A short delay separates each cw and ccw motion
;This is repeated for ever.
;The stepping rate (steps/sec) is set by the delay subroutine
;coilA connected to pa7
;coilB connected to pa6
;coilC connected to pa5
;coilD connected to pa4
;start initialisation
org $e000
cli
; set bit7 of PACTL to output
ldaa #$80
staa $1026 ;bit7=1 in DDRA7
;finish initialisation
;rotate shaft cw by 1 rev (200 steps)
;use the Y index reg as a counter
again ldy #50 ;200steps=50 x 4 where 50 is loaded into IY and
;4=four steps of cycle
revcw jsr cw
dey
bne revcw
;now give short delay before reversing (share the use of the delay subroutine)
ldy #25
sdel1 jsr delay
dey
bne sdel1
;now rotate shaft ccw by 1 rev (200 steps), again use the Y index reg as a counter
ldy #50
evccw jsr ccw
dey
bne evccw
;now give another short delay before reversing again(share the use of the delay
subroutine)
4.6 Stepper Motor Control 73

ldy #25
sdel2 jsr delay
dey
bne sdel2
jmp again
;subroutine to step through 1 cycle of full-step mode in
;clockwise direction
cw ldaa #$80 ;step1 of cycle(coilA=on)
staa $1000
jsr delay

ldaa #$40 ;step2 of cycle(coilB=on)

staa $1000
jsr delay
ldaa #$20 ;step3 of cycle(coilC=on)
staa $1000
jsr delay
ldaa #$10 ;step4 of cycle(coilD=on)
staa $1000
jsr delay
rts
;subroutine to step through 1 cycle of full-step mode in
;counter clockwise direction
ccw ldaa #$20 ;step3 of cycle(coilC=on)
staa $1000
jsr delay
ldaa #$40 ;step2 of cycle(coilB=on)
staa $1000
jsr delay
ldaa #$80 ;step1 of cycle(coilA=on)
staa $1000
jsr delay
ldaa #$10 ;step4 of cycle(coilD=on)
staa $1000
jsr delay
rts
;delay routine works close to pull-in speed of motor
delay ldx #500 ;
del1 dex
bne del1
rts
74 4 Microcontroller

----------------------------------------
Program 4.16: Half Step Mode
----------------------------------------

;Program to drive a 200step/rev unipolar stepper motor in half-step mode

(400steps/rev).
;The motor runs smoother in half-step mode.
;Half-step mode(cw) = coilA then coilsA+B then coilB then coilsB+C
;then coilC then coilsC+D then coilD then coilsD+A then repeat cycle.
;Half-step mode(ccw)= coilD then coilsC+D then coilC then coilsB+C
;then coilB then coilsA+B then coilA then coilsD+A thenrepeat cycle.
;The motor shaft rotates clockwise by 1rev(400steps) then counterclockwise by
;1rev(another 400steps)
;A short delay separates each cw and ccw motion
;This is repeated for ever.
;The stepping rate (steps/sec) is set by the delay subroutine
;coilA connected to pa7
;coilB connected to pa6
;coilC connected to pa5
;coilD connected to pa4
;start initialisation
org $e000
cli
;set bit7 of PACTL to output
ldaa #$80
staa $1026 ;bit7=1 in DDRA7
;finished initialisation
;now rotate shaft cw by 1 rev (400 steps)
;use the Y index reg as a counter
again ldy #50 ;400steps=50 x 8 where 50 is loaded into IY and
;8=eight steps of cycle
rvcw jsr cw
dey
bne rvcw
;now give short delay before reversing (share the use of the delay subroutine)
ldy #100
hdel1 jsr delay
dey
bne hdel1
;now rotate shaft ccw by 1 rev (400 steps)
;again use the Y index reg as a counter
ldy #50
revccw jsr ccw
dey
bne revccw
;now give another short delay before reversing again(share the use of the delay
subroutine)
4.6 Stepper Motor Control 75

ldy #100
hdel2 jsr delay
dey
bne hdel2
jmp again
;subroutine to step through 1 cycle of half-step mode in
;clockwise direction
cw ldaa #$80 ;step1 of cycle(coilA=on)
staa $1000
jsr delay
ldaa #$c0 ;step2 of cycle(coilA+B=on)
staa $1000
jsr delay
ldaa #$40 ;step3 of cycle(coilB=on)
staa $1000
jsr delay
ldaa #$60 ;step4 of cycle(coilB+C=on)
staa $1000
jsr delay
ldaa #$20 ;step5 of cycle(coilC=on)
staa $1000
jsr delay
ldaa #$30 ;step6 of cycle(coilC+D=on)
staa $1000
jsr delay
ldaa #$10 ;step7 of cycle(coilD=on)
staa $1000
jsr delay
ldaa #$90 ;step8 of cycle(coilD+A=on)
staa $1000
jsr delay
rts
;subroutine to step through 1 cycle of full-step mode in
;counter clockwise direction
ccw ldaa #$10 ;step7 of cycle(coilD=on)
staa $1000
jsr delay
ldaa #$30 ;step6 of cycle(coilC+D=on)
staa $1000
jsr delay
ldaa #$20 ;step5 of cycle(coilC=on)
staa $1000
jsr delay
ldaa #$60 ;step4 of cycle(coilB+C=on)
staa $1000
jsr delay
ldaa #$40 ;step3 of cycle(coilB=on)
76 4 Microcontroller

staa $1000
jsr delay
ldaa #$c0 ;step2 of cycle(coilA+B=on)
staa $1000
jsr delay
ldaa #$80 ;step1 of cycle(coilA=on)
staa $1000
jsr delay
ldaa #$90 ;step8 of cycle(coilD+A=on)
staa $1000
jsr delay
rts
;delay routine works close to pull-in speed of motor note that pull-in speed of
;half step mode is approx twice that of full-step mode but note that angular
;velocity of pull-in speed for both modes is approx the same.
delay ldx #250 ;
del1 dex
bne del1
rts

4.7 Serial Communications

We can use a computer to communicate with the microprocessor via RS232 connect
to COM1 of the microprocessor unit. A simple experiment starts by opening the
hyper terminal for the interface. We load a program into the microprocessor after
configuring the serial port of the computer. The serial communication may be set as
follow: 8 data bits; 1 stop bit; no parity bit; no flow control; 9600 baud rate. We
write the program in the high levels language. The function is to send a character
from the PC to the microprocessor. The microprocessor will in turn reply, with a
character. The character is to be input from the PC to the microprocessor through
the hyper terminal. It will receive and return a ‘character+1’. Example, you send an
‘A’ to the microprocessor, the hyper terminal will receive a ‘B’. If you send a ‘1’,
you receive a ‘2’.
4.7 Serial Communications 77

---------------------------------------------------
P4.17 PC Communication With MicroP
---------------------------------------------------
#include “startup.bas”
byte recu_pc
initrs232in()
Do ‘loop here
If recu_pc<>0 Then ‘character from PC
Putcharrs232(recu_pc+1) ‘send back to PC from MicroP
Recu_pc=0
End If
Loop
Function putcharrs232(x) ‘send a character back to PC
Do
loop until SCSR.7=1
SCDR=x
End Function
Function initrs232in() ‘rs232 initialisation
ASM ldx #rs232in ‘activate interrupt rs232
ASM stx $00FE
End Function
‘Character send to microprocessor
ASM rs232in: staa recu_pc
ASM rti
Chapter 5
Electronics Control

5.1 Servo Motor Control

Laboratory experiment for the servo motor is set up to demonstrate and study the
characteristics of the implemented PID controller system. The motor’s controller
output is affected by the backslash. We eliminate the backslash by using the
backslashing gear for the dc motor. The gearbox of the motor can be of any
selection ratio for use. The joystick is used to jack in the input voltage reference
signal to the motor plant. The selectable switches are to select the relevant P, PI, PD
and PID controllers for analyses. We connect the controller output to the motor.
The BNC connector for the servo motor locates at the right-hand side of the
controller box. We can change each of the proportional, integrator and differentiator
parameters by the parameters’ values selection switches. The meters gives the
readout of the analog voltage readings. The actual readout connects to the oscil-
loscope through another BNC connector located at the bottom right of the control
box. From here, we study the transient and steady state characteristics of the motor
plant. Instead of manual input from the joystick, we can adjust the black knob to
input the ranges of frequencies to the servo motor. A square wave signal oscillator
is developed to eliminate the input signal required from the function generator.
Hence, we save space from the occupied function generator (Figs. 5.1 and 5.2).

© Springer Science+Business Media Singapore 2016 79

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_5
80 5 Electronics Control

Fig. 5.1 Servomotor

controller box

Fig. 5.2 Error detecting

circuit

As shown in the error detecting circuit, the voltage follower serves the purpose
of stabilizing the feedback signal from the motor plant. Following the differentiator
circuit to produce the error signal between the voltage signal feedback and the
desired input voltage signal from the square wave generator. This error signal then
feeds back into the PID controller circuit. We observe a gain of 1 in the differen-
tiator circuit. Figure 5.3 illustrates the full control circuit for the motor plant.

5.2 Square Wave Generator

A square wave oscillator provides the input of the controller circuit with a desired
frequency of oscillation to study the motor. The circuit according to Fig. 5.4 is the
design.
Oscillatory Circuit
Desired criterior 0.2 Hz, 1 V output:
Using 12 V d.c. power supply
Choose C = 3.3 μF; R = 0.68 MΩ; R1 = R2 = 1 MΩ; R3 = 1 kΩ; R4 = 28 kΩ
 5.2 Square Wave Generator

Fig. 5.3 Servomotor control block diagram

81
82 5 Electronics Control

Fig. 5.4 Square wave

oscillator circuit

Formula: T ¼ 2RC In½ð2R1 þ R2Þ=R2

¼ 2RC In 3
¼ 4:95 s
Freq ¼ 1=T ¼ 0:2 Hz
ð5:1Þ
Measured value: VA ¼ 28 V
Amplifier Ratio : Ratio ¼ R3=R4
Voltage Output : Vo=p ¼ VA:ðR3=R4Þ
¼ 28ð1=28Þ ¼ 1 V

The VA value is adjustable to 12.64 kΩ to the +12 V side of the potentiometer

to give a 28 V peak to peak signal. The peak to peak voltage output is equivalent to
1 V signal. Alternatively, we can generate square wave by using 555 timers. The
555 timer generates square wave oscillations when operating in astable mode. With
the values of R1, R2 and C given, the frequency will range from a few Hz to several
hundreds of kHz. To provide a low frequency, you can replace the 0.01 μF
capacitor C with a higher value. The formula to calculate the frequency is given by:

TL ¼ 0:693  R2  C ð5:2Þ

TH ¼ 0:693  ðR1 þ R2Þ  C ð5:3Þ

1=f ¼ 0:693  C  ðR1 þ 2  R2Þ ð5:4Þ

The duty cycle is given by:

% duty cycle ¼ 100  ðR1 þ R2Þ=ðR1 þ 2  R2Þ

ð5:5Þ
¼ 50%

For ensuring an approximately 50 % duty ratio, R1 should be small when

compared to R2. But R1 should be no lesser than 1 kΩ. A good choice would be R1
in kilo-ohms and R2 in mega ohms. You can then select C to tune the range of
desired frequencies. With values of R1 = 100 kΩ; R2 = 2 MΩ; C = 1 μF;
RL = 1 kΩ. for Eq. (5.4), the frequency of 0.35 Hz can be found. However,
5.2 Square Wave Generator 83

5v to15v
2

Fig. 5.5 Timer 555 square wave generator

by changing the resistor R1 to 1 kΩ, it produces a more precise square wave of

50 % duty cycle. Thus, R1 is the compromising factor leading to the 50 % duty
cycle, square wave generator. We can connect a load resistor RL across pin 8 and
pin 3 of the timer as a pull-up resistor. By replacing the adjustable R2 resistor with a
diode (IN914) connected in parallel with a fixed R2 (anode of diode to pin7,
cathode to pin6 of timer), we can have a similar effect as a timer oscillator circuit.
The fixed R2 resistor (parallel with the diode), can be of equal value as R1. The
output is equal to the supply voltage (Fig. 5.5).

5.3 PID Controller

The proportional-integral-derivative controller is used widely in the industries. The

direct and easy implementation has made it popular in the modern control system.
Assuming simple amplifiers connected in RC forms as shown in Fig. 5.6. The final
gain for the PID circuit is derived as follow, where we set Kp gain to 1.

G ¼ kpð1=Tis þ Tds þ KpÞ

ð5:6Þ
¼ Ki=s þ KDs þ K; where Ki ¼ kp=Ti; KD ¼ kp  Td

Figure 5.7 shows the full implementation of the analogue PID controller. The
settings are as follow: CDi = 0.1 μF; CIf = 0.22 μF; C4 = 10 nF

R1 ¼ 1:5 kX; R2 ¼ 15 kX; R3 ¼ 4:7 kX; Rkf ¼ 10 kX; Rki ¼ 10 kX;

84 5 Electronics Control

Differentiator Circuit
Gain ratio = - RC*s = -Td*s

Integrator Circuit
Gain ratio = - 1/(RC*s) = - 1/(Ti*s)

Proportional Circuit
Gain ratio = - R2/R1= - kp

Fig. 5.6 Differentiator, integrator, proportional circuits

We define integrator, differentiator and summer corner frequencies as:

fi1 ¼ 0:2 Hz; fi2 ¼ 4 Hz; fd1 ¼ 90 Hz; fd2 ¼ 800 Hz; fs ¼ 1 kHz

Resistor calculated for the summer corner frequency is:

R4 ¼ 1=ð2p  C4  fsÞ ¼ 15:92 kX ðchoose R4 ¼ 16 kXÞ

Proportional d.c gain is, K ¼ Kp  kp ¼ ðRkf  R4Þ=ðRki  R3Þ ¼ 3:404

Suitable resistor for integrator corner frequencies are:

RIf ¼ 1=ð2p  CIf  fi1Þ ¼ 3:617 MX ðchoose RIf ¼ 3:6 MXÞ

RIi ¼ 1=ð2p  CIf  fi2Þ ¼ 0:1809 MX ðchoose RIi ¼ 180 kXÞ

Proportional d.c gain is, Ki ¼ ðRIf  R4Þ=ðRIi  R2Þ ¼ 21:33 Suitable resistor
for differentiator corner frequencies are:

RDf ¼ 1=ð2p  CDi  fd1Þ; RDf ¼ 17:68 kX ðchoose RDf ¼ 18 kXÞ

RDi ¼ 1=ð2p  CDi  fd2Þ; RDi ¼ 1:989 kX ðchoose RDi ¼ 2 kXÞ

Proportional d.c gain is, KD ¼ ðRDf  R4Þ=ðRDi  R1Þ ¼ 96

SðsÞ ¼ R4=C4 ¼ R4=ðR4:C4:s þ 1Þ

5.3 PID Controller 85

16k

4.7k

To motor plant

Fig. 5.7 PID analogue controller

86 5 Electronics Control

Fig. 5.8 Frequency response bode plot

 
RKf R4  SðsÞ RIf R4 SðsÞ s  CDi  RDf R4  SðsÞ
GðsÞ :¼  þ  þ 
RKi R3 RIi  ð1 þ s  RIf  CIf Þ R2 1 þ s  CDi  RDi R1
ð5:7Þ

The response Mag db(w) = 20 log (|G(jw)|); N = 60,000, i = 1 to N Phase angle

(w) = arg(G(jw))180/π; wi = i; fi = wi/2π (Fig. 5.8).
5.4 Control of an Electro-pneumatic Mechanism 87

5.4 Control of an Electro-pneumatic Mechanism

The analogue voltage control circuit measures the differences between the desired
set voltage and the feedback voltage to operate the solenoid valve of a system. The
electronics circuit diagram is as shown in Fig. 5.9. The function of the first
amplifier in the circuit is to calibrate the errors of the transducer signal feedback.
We define this as the zeroing of the voltage feedback to a reference point. Next, we
can set the desired voltage at the terminal of the differential amplifier. The voltage
signal will be fed into the voltage comparator together with the feedback voltage.
A supply voltage of 5 V offsets the differences of the set voltage and the feedback
voltage in the next amplifier circuit. Finally, the output of the analogue control
circuit is fed forward to the solenoid valve of the rodless cylinder to initiate the
linear motion. The electronic circuit of the analogue voltage controller consists of
three types of amplifier circuits as shown above. We can ground the inverting
amplifier using a 1 kΩ, R3 resistor (Fig. 5.10).

R2
For differential amplifier; Vo ¼ ðV2V1Þ ð5:8Þ
R1
R2
For inverting amplifier; Vo ¼  V1 ð5:9Þ
R1
 
R3 R3
For summing amplifier; Vo ¼  V1 þ V2 ð5:10Þ
R1 R2

The analogue signal control circuit output voltages for each of the stages are
derived as follow. For simplicity, we use 1 kΩ value for all the resistors. So the
amplification is straight forward.
 
R3 R3
First Stage: V1 ¼  Vtrim þ Vin ð5:11Þ
R1 R2

R6
Second Stage: V2 ¼  V1 ð5:12Þ
R4

Fig. 5.9 Amplifier circuits a Differential. b Invertor. c Summing

88 5 Electronics Control

Fig. 5.10 Analogue signal control circuit

R8
Third Stage: V3 ¼ ¼ ðVsetV2Þ ð5:13Þ
R7
 
R11 R11
Fourth Stage: V4 ¼  Voffset þ V3 ð5:14Þ
R9 R10
 
R14
Final Stage: Vout ¼  V4 ð5:15Þ
R12

The voltage input to the displacement output of the rodless cylinder mechanism
system is as shown in the plot. The static sensitivity is calculated approximately as
5.625 V/m. There is an offset in the initial system as shown in the graph. That arises
the need for the 20 kΩ potentiometer. The zeroing adjustment of the transducer
potentiometer is to eliminate the error of the feedback voltage, from the displace-
ment transducer when starting. The 20 kΩ referencing potentiometer calibrates the
misalignment of the signal feedback, due to some physical, mechanical error or
degradation of the components which inherits in the system. We can adjust the
potentiometer or trimmer to counteract on the positive or negative transducer signal
feedback. The system is trimmed to 5 kΩ as shown in the circuit. The position of
the adjusted voltage of the trimmer can eliminate the error of the transducer
feedback voltage (as referenced to the static sensitivity graph). For example, when
the set and offset voltages are all set to zero, there is no distance travel or any
displacement. So the voltage from the transducer should feedback a 0 V signal. But
when referring to the static sensitivity graph, there is an offset of 1 V in the system
for zeros displacement. So we trim the voltage trimmer to a negative 1 V same
signal. It eliminates the transducer signal, so the final output of the analogue control
circuit is zero at the beginning. Therefore, there will be no displacement for the
linear system before we inject a desired set voltage (Fig. 5.11).
Also, the voltage offset is set to 5 V for the system because the rodless cylinder
mechanism is referenced and positioned to this fixed distance at the beginning.
5.4 Control of an Electro-pneumatic Mechanism 89

Fig. 5.11 Static sensitivity graph

As shown in the graph, the distance for the 5 V offset will be approximately 89 cm
away from the starting end of the rodless linear mechanism system. Therefore, the
two potentiometers serve as a signal compensation and voltage allowance for the
mechanical system.
Chapter 6
Electrical System

6.1 Elevator Control

PLC control systems are invested heavily in the market to perform many different
control functions. Brands like Hitachi, Omron, Allen Bradley, etcs, are popular in
their usage, serving the various sectors in the industrial. Due to ease of maintenance
and the flexibility in configuration, the PLC control system is used worldwide.
Elevator control employs PLC.

6.2 Programmable Logic Controller

We used the Hitachi E-Series PLC for a large number of sequential control
applications. The ladder logic diagram is programmed and applied in many
applications. The operation of the elevator control for a four storey building is one
example. The elevator or lift system consists of three PLC program structures. They
are the lift up/down control structure, the lift door open/close control structure and
the lift lighting control structure. Figure 6.1 is the lift system layout for the design.
Table 6.1 shows the standard input and output lift functions for the operational
control of the lift system. There are altogether thirteen push buttons, eight limit
switches and seventeen indicating lights in the elevator system. Besides, we have
four solenoid relays to control the lift up/down and door open/close positions.

6.3 Ladder Diagram Control Structures

The ladder logic control design for the lift system consists of three main parts. They
are

© Springer Science+Business Media Singapore 2016 91

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_6
92 6 Electrical System

1 2 3 4
L1 L2 L3 L4 LS1

P1 up L5 LS2
P2 down L6
P3 up L7 LS3
P4 down L8 LS7 LS8
P5 up L9 Open Close
P6 down L10 LS4
LS5
LS6

L1 – 50 P1 – 00 LS1 – 100
L2 -- 51 P2 – 01 LS2 – 101 P7 EMG. L11
STOP
L3 – 52 P3 – 02 LS3 – 102
L4 – 53 P4 – 03 LS4 – 103 P8 4 L12
L5 – 54 P5 – 04 LS5 – 104
L6 – 55 P6 – 05 LS6 – 105 P9 3 L13
L7 – 56 P7 – 06 LS7 – 106
2
L8 – 57 P8 – 07 LS8 – 107 P10 L14
L9 – 150 P9 – 10 COM
1
L10 – 151 P10 – 11 P11 L15
L11 –152 P11 – 12
L12 – 153 P12 – 13
L13 – 154 P13 – 14 P12 P13
L14 – 155 OPEN CLOSE
L15 – 156 Up – 60 Close – 62
L16 – 157 Down – 61 Open -- 63 L16 L17
L17 –160 COM COM
COM
Fig. 6.1 Lift references

1. Part 1: Indicating Lights

2. Part 2: Lift door open/close
3. Part 3: Lift up/down
The structural design makes it much more easy to learn and understand.
Furthermore, lift control designer can just concentrate and change the required part
6.3 Ladder Diagram Control Structures 93

Table 6.1 Elevator I/O I/Os Functions

functions
Inputs P1 to P13 Push button switches
LS1 to LS8 Limit switches
Outputs L1 to L17 Indicator lights
Up Lift motor ‘up’ control
Down Lift motor ‘down’ control
Open Lift door ‘open’ control
Close Lift door ‘close’ control

without worries or concerns of undoing the whole system. The designer can then
combine the three parts as one continuous ladder control diagram for the elevator
system.

6.3.1 Part 1:- Indicating Lights

As shown in Fig. 6.2 is the ladder diagram structure for the lift indicating lights. We
installed these lights outside the lift for indication to passengers waiting outside the
lift. We connect the internal relays 50–53 of the PLC to the indicating lights outside
the elevator. The light indicators from 53 to 50 are from level 4 to level 1
respectively. LS2 activates the 4th storey limit switch, LS3 represents the 3rd storey
limit switch and so on. We used only four limit switches in the design.
Alternatively, we may want to use two limit switches to signal for the lift
indicator (Fig. 6.3). It depends on the design engineer whether to fix one or two

Fig. 6.2 Indicating lights

94 6 Electrical System

Fig. 6.3 Indicating lights alternate design

limit switches for activating the light indicators. Additional limit switches will
ensure a more fully proofed system at the expenses of having more maintenance
and costs. Moreover, additional hardware increases the chances of hardware failure
of the system also.

6.3.2 Part 2:- Lift Door Open/Close

The solenoid relay 63 of Fig. 6.4 energises the lift door open motor while the coil
62 energise the lift door to close. Lamp indicators are 157 and 160. We use the on
delay timer T02, to delay the lift door closing when it is open. The delay is set to
2 s. When the elevator door is opening, pressing the close button P13 will not
activate the lift door to close. On the other hand, pressing the open button P12 will
bypass the door closing to open the lift door. The contacts 60 and 61 are from the
lift up/down motors. We install the on delay timer T04 to delay the timing for
opening the lift door when the elevator reached the level or when the up/down
motor has de-activated. The timer is set to hold for one second before the elevator
door operates to open. So the lift up/down motor will not be activated first (6 s
delay). Another scenario is to press the buttons outside the lift to be able to open the
lift door. The second rung of the ladder diagram of the buttons P1 to P6 hold the lift
door back from closing when the lift stations at the respective level.
6.3 Ladder Diagram Control Structures 95

Fig. 6.4 Lift open/close

96 6 Electrical System

The first line in the elevator open/close ladder diagram ensures the elevator is
stationary before it can open its door. While the lift is moving, the coil 60 or 61 will
be activated to open its contact. Hence, logic line one of the ladder diagram is open
circuit. So the lift door will not be opening. Once reached (after delay 1 s), the lift
door opens until it touches the open limit switch LS7. Conversely, the lift door will
close (if there is no interference) until it activates the close limit switch LS8, to
de-energize the close relay motor 62. Then, the program continues to the lift
up/down ladder diagram to decide about the elevator vertical movement.

6.3.3 Part 3:- Lift Up/Down

The lift up and down moving control system is shown in the ladder diagram of
Fig. 6.5. Beside it are the mnemonics or logic codes for the ladder diagram pro-
gram. The up/down control of the elevator is cut off by the limit switches (from LS2
to LS5) in each of the levels when activated.
The lift control is manually input by pressing any of the six buttons outside the
lift or any of the four buttons inside the lift. Altogether, we have ten press button
input for the elevator control system. The lift up/down movement activates when
LS8 limit switch is closed. Furthermore, it depends on the location level of the
elevator to decide whether to move up or down. That depends on the activation of
the limit switches (LS2 to LS5) in each level. The indicating lights inside the lift,
from L11 to L15 (contactor output 152 to 156) only connect to light indicators. We
do not connect them to any contact point (152 to 156). Whereas, the indicating
lights outside the lift, from L5 to L10 (contactor 54 to 57 and 150 to 151), all are
connected to contactor points and relays. For instance, if we pressed button P9
(level 3) inside the lift, the button P2 or P3 will light up. Which button light up
depends on whether the lift is above or below the level 3. Let us say, the lift is in
level 2 or 1, so P3 (light L7) lights up. If the lift is at level 4, L6 will lit instead of
L7 outside the lift. If we press P2 or P3 outside the lift, the P9 button inside the lift
will light up also. Thus, passengers outside or inside the lift know the elevator will
arrive and stop at the levels where there are lighted buttons. The relays (60 and 61)
each connects to their respective motor for activating the lift up and down. The LS8
limit switch ensures the lift is closed before it can move up/down. We install on
delay timer T05 to T08 to control the elevator from overshot when the passengers
had pressed the lift from different levels. These four timers have a delay of 6 s each,
to cater time for the lift to stop at the activated level before it responses to its next
vertical (up/down) movement in the same direction. Whereas the three seconds
timer T09 and the two seconds timer T11, allow enough time to open the lift door
first before it responses to the next passenger level for the lift to move in the
opposite direction. The delay prevents the lift from responding to move in the
opposite direction when it has reached the passenger level. Timers T09 and T11 are
6.3 Ladder Diagram Control Structures 97

Fig. 6.5 Lift up/down

98 6 Electrical System

Fig. 6.5 (continued)

set differently to prevent both up/down motors of the lift from energizing simul-
taneously. Together with the ladder diagram controlling the lights and the lift door
opening and closing, the elevator functions as a complete working system.

6.4 Safety Control Features

We can add safety features to the lift control system. For example, we can put two
more limit switches to stop the lift at the highest and the lowest level. Limit switch
LS1 is the safety control to halt the elevator from rising when the LS2 limit switch
at level 4 fails. Similarly, LS6 limit switch is mounted to prevent the lift from
descending further if limit switch LS5 fails. Secondly, we implement a timer delay
or on delay timer T01 to activate the lift to rise. The purpose is to delay energizing
the upward lift motor if both the up and down motors activate at the same time
when the lift door (LS8 limit switch) is at closed position. That is cause by both the
upper and lower level passengers energizing the lift buttons when the lift is held
stationary at the middle level (between both the passengers level). So, the lift will
activate the move down motor 61 first. Furthermore, we can install the emergency
button inside the lift too. Button P7 is the emergency latch button to halt the
elevator at the standstill during the emergency. Another implementation is to move
the elevator to the level 1 or ground level when nobody is using. If any of the
6.4 Safety Control Features 99

Fig. 6.6 Lift safety control features

buttons from P1 to P6 and from P8 to P11 are not press, the elevator will auto-
matically lower itself to the ground level (level 1). That is the initial stationary
position (level) of the lift to wait for passengers. Fifth, we can add on delay timer
(T03) to delay the lift from moving up once the door is closed. Lastly, we add a
timer delay, T10 to have different timing for energizing the lift down motor
(61) once the lift is closed. The on delay timer, T03 is set to two seconds while T10
is set to 1 s delay. Thus, the timer T01 can be considered redundant for the situ-
ation. These six additional features of the lift system are as shown in Fig. 6.6.
100 6 Electrical System

Table 6.2 Elevator on-delay T01 T02 T03 T04 T05 T10 T09 T11
timer settings T06
T07
T08
1s 2s 2s 1s 6s 1s 3s 2s

We can include them to replace the last eight rungs of Fig. 6.5 to have the complete
safety elevator system (Table 6.2).
The lift timing control is done by the on delay timers from T01 to T11. The
settings for the timings ensure the elevator operates without fail. It guarantees the
smooth transition of the lift to the next level and the efficiency of the lift control
system. With the presence of the timers T10 and T03 in the system design, we can
lengthen T02 to 4 s delay. The extra delay time holds the lift door longer in the
open standstill position. It caters for the concern of the elderly to enter the lift. At
the same time, we can shorten T05 to T08 to 4 s also. In reducing these timing
delays, enable the lift passenger buttons to be more active. By doing so, we
improved the efficiency and user friendliness of the elevator system.
Chapter 7
Power Flow

7.1 Power System Analysis

Numerical methods are used to solve for complex non-linear problems in power
system analysis. One of the most popular techniques is the approximation of the
non-linear algebraic equations to linearise equations. The Newton-Raphson method
is an iterative algorithm, which only makes use of the first few terms of the Taylor
series function f(x). It is used practically in load flow power network. Power
efficiency is crucial for load flow transmission and distribution. Transmission lines
circuit protections and overloads need to be calculated to monitor and control the
network system. So network design and line regulation are optimized for better
efficiency.

7.2 Newton Raphson Formulation

By the Newton-Raphson formulation, we have the first-order terms from the Taylor
series expansion shown in Eq. (7.1). Where e stands for the amount of error
computed. x0 is the initial guess for the unknown variable. We set f (x0 + e) to zero
and solve for the error in the system (see Eq. (7.2)).

fðx0 þ eÞ ¼ fðx0Þ þ f 0 ðx0Þe ð7:1Þ

e0 ¼ fðx0Þ=f 0 ðx0Þ ð7:2Þ

xn þ 1 ¼ xn þ e n ð7:3Þ

The unknown value for the system is updated as in Eq. (7.3) until the error is zero.

© Springer Science+Business Media Singapore 2016 101

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_7
102 7 Power Flow

We give an illustration of the N-R program written in Pascal language. It

computes two simultaneous equations using Newton-Raphson formulation. The
convergence is accurate. We just need to input two unknown values of the
approximations as in the equations, and the program computes the result by itself
(Fig. 7.1).

--------------------------------------
Program 7.1: Newton Raphson
---------------------------------------
Program Newton_Raphson;

VAR
X1,X2,F1,F2,X11,X22,T:REAL;
J11,J22,J12,J21,IJ11,IJ22,IJ12,IJ21,DET:REAL;
COUNT:INTEGER;
BEGIN
WRITELN;
WRITELN('Newton Raphson Technique for Simultaneous Equations');
WRITELN(' 3 2 2');
WRITELN('F1= 2X1 + 3X1 X2 - X2 - 2 = 0');
WRITELN(' 2 2 2');
WRITELN('F2= X1X2 + 2X1 - 3X2 + 16 = 0');
WRITELN;
WRITELN('ENTER TWO VALUE OF APPROXIMATION');
WRITELN;
READLN(X1,X2);
WRITELN('X1 = ',X1:4:5,' X2 = ',X2:10:5);
WRITELN;
WRITELN;
WRITELN('COUNT':8,'X1':17,'X2':17);
WRITELN(COUNT:8,X1:17:5,X2:17:5);
X11:=1.0;
X22:=1.0;
7.2 Newton Raphson Formulation 103

Fig. 7.1 Result of the N-R computation

T:=0.001;
WHILE (ABS(X11)>T) OR (ABS(X22)>T) DO
BEGIN
F1:=2*EXP(3*LN(X1))+3*SQR(X1)*X2-SQR(X2)-2;
F2:=X1*SQR(X2)+2*SQR(X1)-3*SQR(X2)+16;
J11:=6*SQR(X1)+6*X1*X2;
J12:=3*SQR(X1)-2*X2;
J21:=SQR(X2)+4*X1;
J22:=2*X2*X1-6*X2;
DET:=J11*J22-J21*J12;
IJ11:=J22/DET;
IJ12:=-J12/DET;
IJ21:=-J21/DET;
IJ22:=J11/DET;
X11:=-(IJ11*F1+IJ12*F2);
X22:=-(IJ21*F1+IJ22*F2);
X1:=X1+X11;
X2:=X2+X22;
COUNT:=COUNT+1;
WRITELN(COUNT:8,X1:17:5,X2:17:5);
END;
WRITELN;
WRITELN(' THE ANSWER AFTER ',COUNT-1,' ITERATION IS X1 = ',X1:1:5);
WRITELN(' X2 = ',X2:1:5);
104 7 Power Flow

7.3 Load Flow Analysis Using Newton Raphson

Consider the following two-bus network system. We use the Newton-Raphson

technique to investigate the reactive power compensation and solve the load flow
problem. The Newton-Raphson technique is a powerful tool, which provides
convergent most of the time, comparing to the Gauss-Seidel method. The power
generated assumed positive, and the load is assumed negative. We can compute the
voltages, angles, active and reactive power at the buses. Once the power factor and
system losses are known, we can control the power system operation and
enhancement for further planning of the power system network. The computation is
base on the assumption of a steady-state load flow operation. Where the parameters
of the A.C power line are determined.

X
n
Pi calc ¼ jVi Vk Vik j cosðUik þ dk  diÞ ð7:4Þ
k¼1

X
n
Qi calc ¼  jVi Vk Vik j sinðUik þ dk  diÞ ð7:5Þ
k¼1

DPi ¼ Pi spec  Pi calc ð7:6Þ

DQi ¼ Qi spec  Qi calc ð7:7Þ

2 3
@Picalc @Picalc
6 @di @jVi j 7
J¼6 4 @Qicalc @Qicalc 5
7 ð7:8Þ
@di @jVi j
2 3
  @Picalc @Picalc  
DPi 6 @di @jVi j 7 Ddi
¼4 6 7: ð7:9Þ
DQi @Qicalc @Qicalc 5 DVi
@di @jVi j
   
Ddi 1 DPi
¼ ½J ð7:10Þ
DVi DQi
" #    r
ðr þ 1Þ
di dri Ddi
¼ þ ð7:11Þ
Vi
ðr þ 1Þ Vir DVir

Expression (7.4) and (7.5) shows the calculated active and reactive power at bus i.
The differences between the specifications and the calculated power in Eqs. (7.6)
and (7.7) are the changes in the active and reactive power. The Jacobian matrix of
Eq. (7.8) is the partial derivative of Pi and Qi. Equation (7.9) is the system equation.
By substituting the Jacobian matrix into the system, we derive Eq. (7.9).
7.3 Load Flow Analysis Using Newton Raphson 105

Fig. 7.2 Two-bus power network circuit

By rearrangement, we arrive at Eq. (7.10). Equation (7.11) is the updated value of

the voltages and the angles. These values are then substituted back into the Pi calc
and the Qi calc for another round of iteration. We matched the mismatch equations
ΔPi and ΔQi against a tolerance value. The computation ended when it satisfies the
tolerances specified.
We specify the voltage and angle of a slack bus to be the referencing source. It
makes up for the losses between the generated power and the load power. If we
know the power and the voltages unknown, it is known as the load bus. Another
name for it is the PQ bus. Generator bus is where we already know the real power
and the magnitude of the voltage. We need to find the reactive power and the
voltage angle of the PV bus (Fig. 7.2).

Y11 ¼ Y22 ð7:12Þ

Y12 ¼ Y21 ð7:13Þ

Z12 ¼ 0:0219 þ j 0:0229

ð7:14Þ
¼ 0:031686 x 46:28

Y12 ¼ 21:8  j22:8

¼ 31:35 x  46:28 ð7:15Þ
¼ 31:55 x 133:72

Y/2 ¼ j0:1
ð7:16Þ
¼ 0:1 x 90

Y11 ¼ Yt ð7:17Þ

Yt ¼ Y/2 þ Y12
¼ j0:1 þ 21:8  j22:8
ð7:18Þ
¼ 21:8  j22:7
¼ 31:47 x  46:15

(Fig. 7.3).
106 7 Power Flow

Fig. 7.3 Transmission line admittance circuit

Fig. 7.4 DOS environment (Y-parameters input)

Go to the DOS environment to key in the Y-parameters’ data for calculations.

Hit Ctrl-Z (return) to save the file (Fig. 7.4).
Let us write a program using Pascal to determine the load flow problem for the
following 3 cases.
Case 1: Determine the voltage and its phase angle at bus 2 (V2 ∟2).
Case 2: Determine the rating of the capacitor (Qc) that if connected at bus 2, it
would make V1 = V2 provided that the generator limits are not being
exceeded; Pg = 3.0 PU; 2.5 ≥ Qg ≥ −1.5 PU
Hint: If you exceed the limits, you can use an inductor (QL) at bus 1.
Case 3: Determine the voltage at bus 2 when the load is suddenly switched off
with the capacitor at bus 2 and the inductor at bus 1, remain connected.
For all cases, calculate the active and reactive power generated at bus 1 (Pg and
Qg). Determine the system losses and power factor. Use a tolerance of 0.0001.
Output the results showing all the power and voltages of both buses.
The algorithm for the program written is as shown at the end of the chapter. The
input guessing values for convergence determines the number of iterations with a
given tolerance. The convergence is best for the approximated X2 (radian) value of
zero. For case 1 the limited X2 (radian) value is not exceeding 0.81. For case 3, the
7.3 Load Flow Analysis Using Newton Raphson 107

maximum for X2 (radian) is not exceeding 0.71. The V2 magnitude approximation

can be of any positive value not less than 1. For case 2, the convergent for the NR
algorithm is between 14.33 and −5.7 for Q2 demand value input estimation. The X2
(radian) angular estimation is best at zero.
Figures 7.5, 7.6 and 7.7 proved the results for the computation of the written
program for the 3 cases.

Fig. 7.5 Case 1 output

Fig. 7.6 Case 2 output (without inductor)

108 7 Power Flow

Fig. 7.7 a Case 2 output, b case 2 output, c case 2 output, d case 2 output

For case 2, we need to derive the subject of interest which is the ΔQ and Δd. The
below equations’ formulations result in the matrix equation of interest (Eq. 7.26)
required for the computation. The first result is as shown in Fig. 7.6 when the
inductor is not present to ensure we do not exceed the generator (Qg) limitation
requirements.
    
DP J11 J12 Dd
¼ ð7:19Þ
DQ J21 J22 DV

DP ¼ J11 Dd þ J12 DV ð7:20Þ

DQ ¼ J21 Dd þ J22 DV ð7:21Þ

J11 Dd ¼ DP  J12 DV ð7:22Þ

7.3 Load Flow Analysis Using Newton Raphson 109

Fig. 7.7 (continued)

1 J12
Dd ¼ DP  DV ð7:23Þ
J11 J11
 
DP J12DV
DQ ¼ J21  þ J22DV ð7:24Þ
J11 J11
 
J21 J21J12
DQ ¼ DP þ J22  DV ð7:25Þ
J11 J11
110 7 Power Flow

Fig. 7.8 Case 3 output

  " 1 J12
# 
Dd DP
¼ J11 J11 ð7:26Þ
DQ J21 J22  J21J12 DV
J11 J11

We insert an inductor (QL) at bus 1 because from Fig. 7.6 we had exceeded the
limitations for the QG(PU) value. Figure 7.7 shows the result is within the limits
after the insertion of the inductor at bus 1. When using the generator limits Qg as the
conversion requirement, the active power at bus 2 is out of the range (see Fig. 7.7a).
When using termination tolerances of ΔV22 and ΔX22, the convergence is slightly
out of the range for load active power load (P2D) at bus bar 2 (see Fig. 7.7b). From
the findings, the convergence requirement is the best when using changes in Pi and
Qi as the terminating tolerances. (The active power load on bus 2 is exactly the same
given value when the iteration stops.) (see Fig. 7.7c, d) Figure 7.8 shows the result of
the voltage at bus 2 when the load is suddenly switched off with the capacitor at bus 2
and the inductor at bus 1, remain connected.
7.3 Load Flow Analysis Using Newton Raphson 111

----------------------------------------------------
Program 7.2:Newton Raphson Load Flow
----------------------------------------------------
Program Newton_Raphson_Load_Flow;
VAR
Y,PHI:ARRAY[1..2,1..2] OF REAL;
READFILE : TEXT;
I,K,COUNT,NUMBER :INTEGER; STRIN:STRING;
CP2,CQ2,P2DC,Q2DC,PL,QL,X22,V22,a,b,c:REAL;
PF,CP1,CQ1,P1DC,Q1DC,P2G,Q2G,PGS,QGS:REAL;
P1DS,Q1DS,P2DS,Q2DS,PGC,QGC,AG,Q2C,Q1I,QGT:REAL;
J11,J22,J12,J21,IJ11,IJ22,IJ12,IJ21:REAL;
J1,J2,J3,J4,T,DET,V1,X1,V2,X2,V2C:REAL;
BEGIN
WRITELN('Newton Raphson Technique for Load Flow');
WRITELN(' 2');
WRITELN('P2DC=31.55*V2*COS(133.72-X2) + 21.8V2 ');
WRITELN(' 2');
WRITELN('Q2DC=-31.55*V2*SIN(133.72-X2) + 22.695V2 ');
ASSIGN(READFILE,'C:DATA1.TXT');
RESET(READFILE);
FOR I:=1 TO 2 DO
FOR K:=1 TO 2 DO
READLN(READFILE,Y[I][K],PHI[I][K]);
CLOSE(READFILE);
FOR I:=1 TO 2 DO
FOR K:=1 TO 2 DO
BEGIN
WRITELN('Y',I,K,' =',Y[I,K]:4:2,' PHASE ANGLE ',I,K,' =',PHI[I,K]:5:2);
PHI[I,K]:=PHI[I,K]*PI/180;
END;
NUMBER:=0;
WHILE NUMBER<3 DO
BEGIN
CP2:=1.0;
CQ2:=1.0;
T:=0.0001;
COUNT:=0;
CP1:=0;
CQ1:=0;P1DS:=0;Q1DS:=0;
PGS:=0;P2G:=0;P2DS:=2.4;P1DC:=0;
QGS:=0;Q2G:=0;Q2DS:=1.8;Q1DC:=0;
V1:=1; X1:=0;
IF NUMBER=2 THEN
BEGIN
112 7 Power Flow

Q1DS:=Q1I;
P2DS:=P2DS-P2DS;
Q2DS:=Q2DS-Q2DS+Q2C;
WRITELN('Load Swtched off with Capacitor(bus 2) and Inductor(bus 1) Connected.');
WRITELN;
END;
WRITELN('Please ENTER V2(more than 1) then angle X2(RAD) for approximation');
WRITELN('Please ENTER X2(RAD) not exceeding 0.707 for convergence (best 0)');
READLN(V2,X2); {*not >0.81 for question 1*}
WRITELN(' V2 = ',V2:4:5,' X2(RAD) = ',X2:4:5);
WRITELN; {*not >0.74 for question 3*}
WRITELN('COUNT':12,'V2':10,'X2(RAD)':20);
WHILE (ABS(CP2)>T) OR (ABS(CQ2)>T) DO
BEGIN
P2DC:=Y[2,2]*SQR(V2)*COS(PHI[2,2])+(Y[2,1]*COS(PHI[2,1]+X1-X2)*V2*V1);
Q2DC:=-Y[2,2]*SQR(V2)*SIN(PHI[2,2])-(Y[2,1]*SIN(PHI[2,1]+X1-X2)*V2*V1);
CP2:=-P2DS-P2DC;
CQ2:=-Q2DS-Q2DC;
J11:=Y[2,1]*V1*V2*SIN(PHI[2,1]+X1-X2);
J12:=Y[2,1]*COS(PHI[2,1]+X1-X2)*V1+Y[2,2]*2*V2*COS(PHI[2,2]);
J21:=Y[2,1]*V1*V2*COS(PHI[2,1]+X1-X2);
J22:=-Y[2,1]*V1*SIN(PHI[2,1]+X1-X2)-Y[2,2]*2*V2*SIN(PHI[2,2]);
DET:=J11*J22-J21*J12;
IJ11:=J22/DET;
IJ12:=-J12/DET;
IJ21:=-J21/DET;
IJ22:=J11/DET;
X22:=(IJ11*CP2+IJ12*CQ2);
V22:=(IJ21*CP2+IJ22*CQ2);
X2:=X2+X22;
V2:=V2+V22;
COUNT:=COUNT+1;
WRITELN(COUNT:8,V2:17:5,X2:17:5);
END;
PGC:=SQR(V1)*Y[1,1]*COS(PHI[1,1])+V1*V2*Y[1,2]*COS(PHI[1,2]+X2-X1);
QGC:=-SQR(V1)*Y[1,1]*SIN(PHI[1,1])-V1*V2*Y[1,2]*SIN(PHI[1,2]+X2-X1);
IF NUMBER=2 THEN
QGC:=QGC-(-Q1I); {Pt=Pg-Pload}
PGS:=PGC+CP1;
QGS:=QGC+CQ1;
X2:=X2*180/PI;
PL:=PGC-P2DS-P1DS;
QL:=QGC-Q2DS-Q1DS;
PF:=PGC/SQRT(SQR(PGC)+SQR(QGC));
AG:=arctan(QGC/PGC)*180/PI;
7.3 Load Flow Analysis Using Newton Raphson 113

WRITELN('AG= ',AG:2:4);
IF AG>0.00 THEN
STRIN:='LAG'
ELSE
STRIN:='LEAD';
WRITELN(' THE ANSWER AFTER ',COUNT,' ITERATION IS : ');
WRITELN('BUS NO VOLT(PU) ANGLE(DEG) PD(PU) QD(PU)
PG(PU) QG(PU)');
WRITELN('------ ------- ---------- ------ ------ ------ ------');
WRITELN('1',V1:13:4,X1:12:4,P1DS:10:4,Q1DS:9:4,PGC:9:4,QGC:9:4);
WRITELN('2',V2:13:4,X2:12:4,-P2DC:10:4,-Q2DC:9:4,'0.0000':9,'0.0000':9);
WRITELN('PLOSSES=',PL:2:4,' PU, ','QLOSSES=',QL:4:4,'
PU,','P.F=',PF:2:4,' ',STRIN:4);
WRITELN('---------------------------------------------------------------------');
NUMBER:=NUMBER+1;
IF NUMBER=1 THEN
BEGIN
WRITELN;
WRITELN('Find Capacitor(bus 2) with V1=V2.');
WRITELN;
WRITELN('Please enter Q2DC then angle X2(RAD) for approximation');
WRITELN('Enter Q2DC value of between 14.33 to -5.7 for convergence');
READLN(Q2DC,X2);
WRITELN(' Q2DC = ',Q2DC:4:5,' X2(RAD) = ',X2:4:5);
WRITELN('COUNT':10,'Q2DC':8,'X2(RAD)':17,'V2C':14,'Pg':14);
CP2:=1.0; Q1I:=1; Q1DS:=Q1I; V22:=1;X22:=1;
CQ2:=1.0;COUNT:=0; QGC:=5.0;PGC:=5.0;V2:=1;
WHILE (ABS(CP2)>T) OR (ABS(CQ2)>T) DO
BEGIN
{*Q2DC:=-Y[2,2]*SQR(V2C)*SIN(PHI[2,2])-(Y[2,1]*SIN(PHI[2,1]+X1-
X2)*V2C*V1)*}
a:=-Y[2,2]*SIN(PHI[2,2]);
b:=-Y[2,1]*SIN(PHI[2,1]+X1-X2)*V1;
c:=-Q2DC;
V2C:=(-b+sqrt(sqr(b)-(4*a*c)))/(2*a);
{ IF V2C>0 THEN
V2C:=V2C
ELSE
V2C:=(-b-sqrt(sqr(b)-(4*a*c)))/(2*a);}
{IF V2C1}
P2DC:=Y[2,2]*SQR(V2C)*COS(PHI[2,2])+(Y[2,1]*COS(PHI[2,1]+X1-
X2)*V2C*V1);
CP2:=-P2DS-P2DC;
{CQ2:=-Q2DS-Q2DC;}
V22:=V2-V2C;
114 7 Power Flow

J11:=Y[2,1]*V1*V2C*SIN(PHI[2,1]+X1-X2);
J12:=Y[2,1]*COS(PHI[2,1]+X1-X2)*V1+Y[2,2]*2*V2C*COS(PHI[2,2]);
J21:=Y[2,1]*V1*V2C*COS(PHI[2,1]+X1-X2);
J22:=-Y[2,1]*V1*SIN(PHI[2,1]+X1-X2)-Y[2,2]*2*V2C*SIN(PHI[2,2]);
J1:=1/J11;
J2:=-J12/J11;
J3:=J21/J11;
J4:=J22-((J21*J12)/J11);
X22:=(J1*CP2+J2*V22);
CQ2:=(J3*CP2+J4*V22);
X2:=X2+X22;
Q2DC:=Q2DC+CQ2; {Q2DC:=-Q2DS-CQ2;}
PGC:=SQR(V1)*Y[1,1]*COS(PHI[1,1])+V1*V2C*Y[1,2]*COS(PHI[1,2]+X2-X1);
QGT:=-SQR(V1)*Y[1,1]*SIN(PHI[1,1])-V1*V2C*Y[1,2]*SIN(PHI[1,2]+X2-X1);
QGC:=QGT-(-Q1DS); {Pt=Pg-Pload)}
COUNT:=COUNT+1;
WRITELN(COUNT:6,Q2DC:15:5,X2:14:5,V2C:14:5,PGC:14:5);
END;
X2:=X2*180/PI;
PL:=PGC-(-P2DC)-P1DS;
QL:=QGC-(-Q2DC)-Q1DS;
PF:=PGC/SQRT(SQR(PGC)+SQR(QGC));
AG:=arctan(QGC/PGC)*180/PI;
WRITELN('AG= ',AG:2:4);
IF AG>0.00 THEN
STRIN:='LAG'
ELSE
STRIN:='LEAD';
Q2C:=-Q2DC-Q2DS;
WRITELN(' THE ANSWER AFTER ',COUNT,' ITERATION IS : ');
WRITELN('BUS NO VOLT(PU) ANGLE(DEG) PD(PU) QD(PU) PG(PU)
QG(PU)');
WRITELN('------ ------- ---------- ------ ------ ------ ------');
WRITELN('1',V1:13:4,X1:12:4,P1DS:10:4,Q1DS:9:4,PGC:9:4,QGC:9:4);
WRITELN('2',V2:13:4,X2:12:4,-P2DC:10:4,-Q2DC:9:4,'0.0000':9,'0.0000':9);
WRITELN('PLOSSES=',PL:2:4,' PU, ','QLOSSES=',QL:4:4,'
PU,','P.F=',PF:2:4,' ',STRIN:4);
WRITELN('---------------------------------------------------------------------');
WRITELN('THE RATING OF THE CAPACITOR IS -Q2DC-Q2DS = ',Q2C:4:4);
WRITELN;
END;
NUMBER:=NUMBER+1;
END;
END.
Chapter 8
Process Control

8.1 Water Tank Control System

The experiment we conduct studies the control response of the water level control
process in the system. We control the water level in the tank for the process control
system. Conventional PID control characteristics demonstrate the dynamics of the
system process. In most parts of the experiment, we only make use of one tank. So
we derived a first order equation G(s) for the main process. The apparatus as shown
in the diagram consists of the double tank system. A separation valve (A) locates
between the tanks. We fix the level and flow sensors on each of the tanks. We
supply a voltage input from zero to 10 V to each different tank. The supply voltage
is proportional to the water level and flow rate of each different tank. A plotter
connects to plot the control characteristics of the system (Figs. 8.1, 8.2, 8.3).
The error signal shows the differences between the set-point and the output. We
first study the proportional control system. By adding a PI controller into the
closed-loop system, we can derive the steady-state level output from the block
diagram with a step input.

8.1.1 First-Order Derivation

Q0 ¼ H=R ð8:1Þ

dh
C ¼ Qi  Q0 ð8:2Þ
dt

© Springer Science+Business Media Singapore 2016 115

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_8
116 8 Process Control

Fig. 8.1 Single tank diagram

Fig. 8.2 System block Controller Coupled Tank

diagram e(s) u(s)
Vi(s) Vo(s)
Set-point error Output

By substitution:

dh H
Qi ¼ C þ ð8:3Þ
dt R

Therefore:



H dh H
Q0 =Qi ¼ = C þ ¼ 1=ðTs þ 1Þ ð8:4Þ
R dt R

By using the controller Kp alone, the closed-loop transfer function is:

C:L T:F ¼ Kp=ðTs þ 1 þ KpÞ ð8:5Þ

The steady-state output level due to a unit step input is (Fig. 8.5):
9
When Kp ¼ 3 : 3=ðTs þ 4Þ ¼ 0:75 Vðplot showing 0:85 V) =
When Kp ¼ 6 : 6=ðTs þ 7Þ ¼ 0:85 V(plot showing 1 V)
;
When Kp ¼ 9 : 9=ðTs þ 10Þ ¼ 0:9 Vðplot showing 1 VÞ

By using PI controller:

C:LT:F ¼ ðKp þ Ki=sÞ=½Ts þ 1 þ ðKp þ Ki=sÞ ð8:6Þ

When limit s → 0, Transfer function equals Ki/Ki.

The steady-state output level due to unit step input is reduced to 1 V.
8.1 Water Tank Control System 117

TANK A TANK B

Water Level
Calibration Output Level
Output Level
Scales Sensor
Sensor
Output Flow
Output Flow Sensor
Sensor
Valve A
Variable DC
Variable DC
0 to +10v
0 to +10v
control
control
voltage
voltage

Common Earth Common Earth

Valve B Valve C

Fig. 8.3 Water tank apparatus

If Kp is fixed at 5 V (Fig. 8.6):

9
When Ki ¼ 0:05 : 0:05=0:05 ¼ 1 V =
When Ki ¼ 1 : 1=1 ¼ 1 V plot is 1 V
;
When Ki ¼ 5 : 5=5 ¼ 1 V
118 8 Process Control

If Ki is fixed at 0.5 V (Fig. 8.7):

9
When Kp ¼ 2:5 : 0:5=0:5 ¼ 1 V =
When Kp ¼ 5 : 0:5=0:5 ¼ 1 V plot is 1 V
;
When Kp ¼ 7:5 : 0:5=0:5 ¼ 1 V

8.2 Single Tank Control

The actual response due to a step input is as shown in the plots. The time constant
we plotted in Fig. 8.4 shows 1.75 min (105 s). For the proportional controller
alone, the steady-state error reduces as Kp increases. We manually offset the input,
to have a clear readout for the differences in the steady-state output plotted. The
actual plot might have slight variation from the derivations for the output level. It is
partly due to the pump’s overhead loss or the plotter pen mechanism loose.
Furthermore, small particles present in the liquid might block the pump suction
head. For a small unit input of 1 V, all these might be significant to cause the output
deviation. The disadvantage of using the proportional Kp term alone is that it
introduces an offset in the final value as shown in the Fig. 8.5. So we introduce the
proportional-integral PI term. It is clearly seen by derivation and from the experi-
ment plot also, that the steady-state error is being reduced to zero. The PI controller
is properly selected depending on the required response by choosing the Kp term
first then the Ki term. As shown in the graphs plotted, we found that the system
exhibits second-order characteristics from the PI controller. With a fixed Kp of 5,
the overshoot gets larger as Ki increases. We are unable to fix the Ki first then
adjust the Kp term. Figure 8.7 depicts the reasons behind. Ringing oscillations are
introduced into the system when Kp is small. Moreover, there is always an

Fig. 8.4 RC time constant

8.2 Single Tank Control 119

Fig. 8.5 Proportional alone

Fig. 8.6 PI output (fixed proportional)

Fig. 8.7 PI output (fixed integral)

120 8 Process Control

overshoot present in the system, which we are unable to eliminate. Thus, Fig. 8.6
shows the correct tuning method to adjust the steady-state final output level without
exhibiting any overshoot. From the PI controller, we can add on an additional
derivative term into the system. The PID controller can be introduced to improve
the system characteristics further. By adding the derivative term, it increases the
overshoot, at the same time reduces oscillations and ringings. Moreover, it also
reduces the time constant at the transient stage. Therefore, it improves the speed of
responses to reach steady-state. So, we learned that the derivative term is used to
compromise the integral term in the PID controller. It enables three selectable
tunings in the PID controller, for a more precise control of the system.
Chapter 9
Machine Learning

Global competition, changing customer’s needs, increasing product reliability and

complexity, volatile economic conditions and higher customer expectations are the
changing set of business requirements that causes the need to optimize productivity,
improving quality and reducing cost. The importance of neural network imple-
mented is to boost the process methodology. We monitor the process parameters in
the record flow chart plotted to analyse and study the progress. Next, suitable
controllers can be implemented. Automatic corrections will be feedback to the
system for the final desired output. Likewise, the process visualization for the plant
can be simulated by software in real time.

9.1 Neural Network in Process Control

A neural network learns your process by observation and adapts its strategy for a
more precise control. Neural networks are self-tuning systems that automatically
take advantages of process upgrades and compensate for system degradation.
Neurocontrol comes into existence to the problem solving of tuning a noisy,
non-linear and complex system. Also, it can reduce the cost of implementing
solutions for the problem. The nonlinear and multivariable capabilities of neural
networks make the technology ideal for direct process control [24] (Fig. 9.1).
The artificial neural network has successfully used in many process control
applications. It allows complexity control and critical monitoring of the process
plant and sensors. In many systems, performance degrades over time due to dete-
rioration of the system components. For compensation, operational parameters are
dynamically tuned to optimize system performance. An ANN can be used to make
decisions about the system operation and adjust the appropriate control to keep the
process operates with optimal efficiency. An advantage of ANN over the traditional
adaptive controllers is that we can continuously update the ANN with new

© Springer Science+Business Media Singapore 2016 121

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_9
122 9 Machine Learning

Fig. 9.1 Neural network process control

information by using a dynamic learning approach. The backpropagation algorithm

is commonly used to train the ANN with the training data which composed of
historical data about the process.

9.2 The Artificial Neurons

The neural network comprises of artificial neurons which are group into three main
layers: input, output and hidden layers. Its function is by the rules of mathematical
equations built into a semantic rule-based implementing as a controller to navigate
the system and monitor its progress through its self-iterative learning process.
The information processing performed may be taken as the signals appearing at
the unit’s input. Or the action potentials to the synapses. The effect (PSP) of each
signal can be approximated by multiplying the signal with some number or weight
to indicate the strength of the synapse. The weighted signals are now summed to
produce an overall unit activation. The unit will produce an output response when
the activation exceeds a certain threshold. This functionality captured in the arti-
ficial neuron the Threshold Logic Unit (TLU) (Fig. 9.2).

A ¼ W1X1 þ W2X2 þ W3X3 þ   

Xn
ð9:1Þ
A¼ WiXi
i¼1

Y ¼ 1 if A  h
ð9:2Þ
Y ¼ 0 if A\h

We suppose there are n inputs with signals X1, X2, etc. The signals have a
boolean output that is ‘0’ or ‘1’. The threshold value is assumed zero. We illustrate
the g value in Fig. 9.3.
9.2 The Artificial Neurons 123

Fig. 9.2 Neural network

Fig. 9.3 Activation graph

The graph shows the difference between the two g values. The g value is higher
when the slope is steeper. For calculation, the g value is always put to one. The A
value is always negative.
Thus the equation:

1
fðxÞ ¼ ^ ðgðAhÞÞ
1þe
ð9:3Þ
1
¼ ðWhere h ¼ 0 and g ¼ 1Þ
1 þ e^ A

9.3 Techniques Involved in the Controllers

Generally, there are five basic approaches in neuro-control techniques. They are:
(1) Supervised control
(2) Direct inverse control
(3) Neural MRAC-type adaptive control
(4) Reinforcement learning
(5) Unsupervised control.
124 9 Machine Learning

(1) Supervised control: We make use of a training set consisting of X(t), and u*(t)
where u* is the target action vectors. The system records the actions as well as
the sensor inputs so that, we can track the responses. We need to calculate the
deviation error for the weight adaptation.
(2) Direct inverse control: It is according to supervised learning. The states of the
system make up the input of the plant, and the targets are the actuating signals.
It is suitable for robot control.

X1; X2 ¼ f ðQ1; Q2Þ

ð9:4Þ
Q1; Q2 ¼ f 1 ðX1; X2Þ

(3) Neural model reference adaptive control (MRAC): A reference model is

supplied to output the desired trajectory. The system is linear, and the
parameters of the plant are not given or predicted. The technique is near to
direct inverse control. Specified equation:
X
U ¼ 0:5 ðXiðtÞ  Xi  ðtÞÞ^ 2 ð9:5Þ
t;i

(4) Reinforcement learning: The output feedback of the plant either aids or
negates the control states depending on the environment reaction. The aiding
signal reinforces those states that contribute to improvement, while the
deducting signal, reduces the states that produce the improper behaviour. It
uses a ‘fuzzy’ approach to the calculated output. The weight changing stops
when the output stabilises in the range of the ‘fuzzy’ categorized output.
(5) Unsupervised control: The output of the subsets are put together at random or
unsupervised, but the members of the same subset are one of its kinds. Thus, it
is suitable for feature mappings based on the similarity of the input patterns.
From all of the above techniques, supervised control is used for neural network
process control system as it enables on-line tuning, changing and process opti-
mization in the system.

9.4 NN Learning Rules

(1) Hebb learning: If the two neurons take the same state at the same time (both
inactive or both active), the output of the weight connection between them
increases (Fig. 9.4).
(2) The perceptron learning rule: We calculate the output error from the deviation
between the desired and actual output. Then, it is used to calculate the changes
in weight in the connectism neurons (Fig. 9.5).
9.4 NN Learning Rules 125

decrease

Fig. 9.4 Hebb learning

Fig. 9.5 Perceptron learning

D weight ¼ o/p i  deviation o=p

ð9:6Þ
New weight ¼ old weight þ D weight

(3) The delta learning rule: It makes use of the same concept as the perceptron
learning, except we divide the deviation errors by the number of the prepro-
cess neurons (Fig. 9.6).

deviation o=p ¼ desired o=p  o=p

D weight ¼ i=p neurons  ððdeviation o=pÞ=ðnumber of i=p neuronsÞÞ
new weight ¼ old weight þ D weight
ð9:7Þ

(4) Backpropagation learning rule: It uses the same tactic as the delta learning
rule, but it allows the changing of weights in the additional hidden layers
introduced. The derivative of the activation function is the sensitivity of the

Fig. 9.6 Delta learning

126 9 Machine Learning

activation function. For an output neuron, it is back propagated to the pre-

decessor hidden neurons to sum up the interconnected weight error. The
derivatives are used to calculate the actual weight changes for both the output
and hidden neurons by multiplying the errors in it.

Derivative of activation function: a0 ¼ a ð1aÞ ð9:8Þ

For all the above learning rules, backpropagation is most popular for process
control as it enables a more precise manipulation such that we can change the
weights in the hidden neurons.

9.5 Selection of the Learning Algorithms

There are two main types of neural nets:

(1) The feedforward neural net
(2) The feedback neural net
(1a) The perceptron net (feedforward net): Only the weights at the output of
hidden neurons are trainable. This layer uses the perceptron linear threshold
activation function. The specified equation is (Fig. 9.7).

New weight ¼ old weight þ ðo=p of predecessor neuron  target deviation

 learning rateÞ
ð9:9Þ

hidden layers

variable weights output layers

fixed weights
input layers

Fig. 9.7 Perceptron network

9.5 Selection of the Learning Algorithms 127

h i j

Fig. 9.8 Backpropagation network

The hidden layer in the system lies between the input and output layer by the
interconnected neurons. There is a direction for the computation of infor-
mation. The number of hidden neurons may be 50 % of the number of input
neurons. It is not always the case to have the same number of input neurons
as the output neurons in the last layer.
(1b) Backpropagation (feedforward net): It allows weights into and output of the
hidden neurons to change during learning. There are sensings of activation
neurons due to weight changes. There are two calculations involved. First, is
to calculate the net output and the error in each neuron in the forward
direction and secondly, to backpropagate the net errors to the preceding
neurons. Besides, backpropagated errors to neurons in predecessor layer are
held at the same time. Thus, the calculated net output formed is slow.
Assuming, neuron i is the preceding of neuron j (Fig. 9.8).
Every unit of the neuron inherits with a non-linear function called the activation
function or the standard sigmoidal function.

1
Equation: fðnetÞ ¼ ^ net
X 1 þ e
where: netj ¼ wij oi ð9:10Þ
i
& fi0 ðnetÞ ¼ oi ð1  oi Þ

We use the data training set to train the system for error reduction. At the same
time, we use the validation set to validate the data. The optimal point, where the
intersection occurs between the validation and training errors, is the point to stop
training. This is the best generalization before the validation error starts to increase
(Fig. 9.9).
128 9 Machine Learning

Fig. 9.9 Best generalisation graph [22]

We calculate the error function by the sum of squared error differences between
the actual and targeted output value.

X
N
E ¼ 0:5 ðt j  y j Þ2 ð9:11Þ
j¼1

Dwij ¼ / r j fj0 ðnetÞ ðt j  y j Þxi ð9:12Þ

where / represents the multiplier factor, r is the learning rate, f(net) the activation
function, x defines the input to the neuron. T is the target output and y is the actual
output. We can assume the factor / to be 1.

d j ¼ r j fj0 ðnetÞ ðt j  y j Þ ð9:13Þ

Dwij ¼ / d j xi ð9:14Þ

Dwih ¼ / di xh ð9:15Þ
X
di ¼ ri fi0 ðnetÞ d j wij ð9:16Þ
j2Ii

where Ii is the set of nodes after the hidden node i. It is known as the fan-out of i.
(1c) Radial Basis Function (feedforward net): The RBF have unique identity such
that they have only one hidden layer with radial basis functions Ψ(x) and
9.5 Selection of the Learning Algorithms 129

produces only linear output. As such, training of the net consists of the
unsupervised part to define the center of the radial basis function and
supervised part to learn the weight. A disadvantage is that we often used it
for the small number of inputs. However, for a large input multi-layer per-
ceptron network, gives a better generalisation.
(2a) Hopfield Net (feedback net): The difference is that weight calculation leads
to only the change of neuron states (pattern learning). The minimal problem
should be modelled so optimizing the objective function minimised the
energy function and that non-fulfillment of a constraint, leads to an energy
increase. Its function is more suitable for pattern recognition. The equation
shows the error function. Where Qi is the neuron input.
X X
Equation:  0:5 þ active neurons þ Qi ð9:17Þ

(2b) Simulated Annealing (feedback net): Another name for it is the Boltzmann
machine. It uses the same principal as Hopfield Net and is a further
enhancement of it to counteract the disadvantages of this system. That is, it
allows the neurons to change state to avoid local minimal of the system. It
consists of visible and hidden neurons interconnected. This system stops at
global minimum. However, the main drawback of the system is that it is slow
and is used for travelling salesman problem.

9.6 The Network Topology

The neural network data processor consists of the input, hidden and output layers.
The decision for the numbers of input neurons and the layers of the hidden neurons
involved depends on the complexity of the system process. The data training set for
the input neurons rely on how we control the system and the input sensors involved.
The neuron numbers in the hidden layers and the number of hidden layers in the
system plant will affect the system monitoring and accuracy predetermined from the
neural processor. Thus, a careful determination of the number of neurons in the
hidden layers are often chosen. The final part is the target monitoring of the plant. It
consists of the last layer of the output neurons. The numbers of output neurons
depend on how many and what kind of action output we are controlling. It is in term
compared with the desired target value, and the deviation is feedback to the pre-
decessor to re-calculate the output to be maintained at the target requirements.
Therefore, a network topology is derived from a given system to be controlled.
130 9 Machine Learning

9.7 MLP Backpropagation Network for Process Control

The neural network based controller and predictor are computerised systems, which
perform optimization in the continuous chemical processes. It monitors the
chemical products and adjusts the effect to produce the desired chemical reactions
and properties in a plant. Most of the existing chemical plant do not have chemical
sensors that can measure its properties. Human interventions are needed to examine
the samples off-line and consistently adjust the plant operating parameters to suit
the target. The neural based system can replace the human intervention to estimate
the chemical properties of the final product on-line automatically. The SCADA
system connected allows the display of the plant information. Human operators can
then make decisions to take further action to counteract conditions such as
degradation of the hardware sensors and tunings of plant parameters to maintain the
desired target performance. An example is a chemical process in a mixer plant. We
can use the system for forecasting as the output produced is the future value of the
input signal (Figs. 9.10, 9.11, 9.12).
Based on backpropagation calculations the following data set were being tab-
ulated for round one of the backpropagation learning:

Sensor & Valve

Heat Exchanger Tank

Fig. 9.10 Mixer plant neural net control system

9.7 MLP Backpropagation Network for Process Control 131

Fig. 9.11 Neural net system processor

Fig. 9.12 Neural network external structure

132 9 Machine Learning

gin4 = 1*0.3 + 0.5*0.5 + 0.2*0.1 = 0.57 gin5 = 1*0.1 = 0.1

act4 = = 0.639 act5 = = 0.525

out4 = 1 out5 = 1

gin6 = 1*1 = 1 gin7 = 1*0 = 0

act6 = 0.73 act7 = 0.5
out6 = 1 out7 = 1

D5 = [0.525*(1 – 0.525)]*(1 – 1) = 0
D6 = [0.73*(1 – 0.73)]*(0 – 1) = – 0.197
D7 = [0.5*(1 – 0.5)]*(1 – 1) = 0
D4 = 0.639*(1 – 0.639)*[0.1*0 + 1*( – 0.197) + (0*0) = – 0.045

W54 = 0*1 = 0 W41 = – 0.045*1 = – 0.045

W64 = – 0.197*1 = – 0.197 W42 = – 0.045*0.5 = – 0.0225
W74 = 0*1 = 0 W43 = – 0.045*0.2 = – 0.009

W54 = 0.1+0 = 0.1 W41 = 0.3 – 0.045 = 0.255

W64 = 1 – 0.197 = 0.803 W42 = 0.5 – 0.0225 = 0.4775
W74 = 0+0 = 0 W43 = 0.1 – 0.009 = 0.091

The Table 9.1 weights tabulation is in accordance to Fig. 9.11. That is, it
depends only on one hidden neural processor. Computation begins until stabilisa-
tion once the three adjustable parameters for the three valve sensors are feedback
into the neural network input. The neural network shows the final values during
stabilization at the seventh round (1 epoch meaning the NN completes 1 cycle for
the total nos. of training examples). So the desired output is also achieved. We
monitor the progress and outcome at the control station. The system shows the
on-line tuning using backpropagation technique to match the desired output of the
plant.
9.8 Chemical Plant NN Feedback Control System 133

Table 9.1 Weights tabulation

0 1 2 3 4 5 6 7
W41 0.3 0.2546 0.2144 0.1823 0.1614 0.1543 0.1622 0.1622
W42 0.5 0.4773 0.4572 0.44115 0.4307 0.42715 0.4311 0.4311
W43 0.1 0.09092 0.08288 0.07646 0.07228 0.07086 0.0724 0.0724
W54 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1
W64 1 0.8034 0.5898 0.3603 0.1182 −0.1309 −0.3798 −0.3798
W74 0 0 0 0 0 0 0 0
D5 0 0 0 0 0 0 0 0
D6 0 −0.1966 −0.2136 −0.2295 −0.2421 −0.2491 −0.2489 0
D7 0 0 0 0 0 0 0 0
D4 0 −0.0454 −0.0402 −0.0321 −0.0209 −0.0071 0.0079 0
O4 1 1 1 1 1 1 1 1
O5 1 1 1 1 1 1 1 1
O6 1 1 1 1 1 1 0 0
O7 1 1 1 1 1 1 1 1
I1 1 1 1 1 1 1 1 1
I2 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5
I3 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2
W Weight; D Deviation; O Output; I Input
Continuous neurons are used for the input values
I1 Quality of catalyst being added to the reactant being encoded to the continuous value of 0–1
with respect to position of the valve
I2 Reactance input to the tank being encoded to the range of 0–1
I3 Range of heat measurement being encoded to the range of 0–1 (from 0 to 100° respectively)
Binary neurons are used for the target output
O5 Catalyst density is set to range of 1–5 units (binary 0)
Catalyst density is set to range of 6–10 units (binary 1)
Target at above 5 units (binary 1)
O6 Temperature is set to the range of 0–35° (binary 0)
Temperature is set to range of 36° and above (binary 1)
Target at below 36° (binary 0)
O7 Reactance quantity finalized to the range of 2–3 at binary 1

9.8 Chemical Plant NN Feedback Control System

The reaction states of the catalyst and the reactance were difficult to analyse.
Instead, their properties and characteristics were collected from laboratory data
analyst. Temperature input is sensed using the temperature sensor. Besides setting
the chemical target quality, the catalyst output density and the reactance quantity
are also set according to the control temperature. However, there arise the difficulty
in determining the parameters of the non-linear system. So we employ the neural
network system to control the variables for the catalyst, reactance and the tem-
perature, to reach its desired states. The neural network can counteract the system
error due to the hardware problem of the mechanical valves or sensors. The weights
134 9 Machine Learning

adapt automatically to the targeted set-point even if the input jams at a position due
to failure. The signal from the neural network feeds into the SCADA (Supervisory
Control And Data Acquisition) system and monitor accurately. The neural network
can also control their output to their final states by using PID controller to adjust the
final control element for the catalyst and reactance. We can include the network
advisor to create alarm and leads to human intervention when the controlled outputs
were absurd. The system thus acts as a safeguard to monitor the on-line process and
estimates the final product with references to the desired requirements. It serves as a
simulation tool also, to test different operating process conditions. It is particularly
advantages when justifying process enhancements and modifications. The final
result will be the product with the desired chemical properties. However, the speed,
number of iterations, accuracy and convergence for the neural network depend on
the careful design of the neurons used. The output of the desired state can also be in
real neurons to control their exact value instead of binary neurons to control a
certain range. System lag time contributes slightly to the drawback of the system.
The delay time to activate the final control element from the neural network or vice
versa when sensing creates an interval to real time reaction. However, in water
process control, system lag time is not critical as we do not expect a fast response
from liquid flow/level turbulence and stabilization.

9.8.1 Process Design

From the case above, we can design a feedback with a sampling period, for
example, one sample per 20 s, to input the detected parameter values into the neural
network system. At the same time, we can also set an integrated delay of 10 s to
output from the NN controller to the plant. The delay allows for the training and
adaptation of the neural network computation. The neural network only activates
once it detects the set of input values. It then goes through the neural network layers
to train for the desired output. We initialize the weights to random values, and
neglect the thresholds in the NN system. Then we perform an off-line simulation to
compute the desired preset NN values for the on-line system. That will reduce the
online adaptation training time of the neural network system. The output of the
neural network is sent to control the valves, after going through the delay. The
system forms a closed loop control for each sampling cycle of the NN control
system.
The Rosemount wireless flow, level and temperature transmitters [28] are
deployed at the field site to do the job. The control station will receive the wireless
transmission signals for the three parameters feedbacked. The closed loop system
performs sampling at 0.05 Hz for the neural network computation. After going
through a delay sampler, the three output valve parameters are feedback instantly
into the field controller through wireless transmission detector. Each of the detected
signals goes through a signal conditioner circuit to scale the signal suitable for the
actuating element. We can also place the signal converter at the NN station before
9.8 Chemical Plant NN Feedback Control System 135

transmission. The output of the scaled signal is fed into the PID controller to adjust
the control valve in real time. After twenty seconds, the flow/level/heat of the
reactance and catalyst will be feedbacked into the neural system wirelessly. Before
that, we can calibrate each of the flow/level/temperature transmitters with their
valves’ positionings. Once calibrated, every valves’ positions will correspond to
their respected variables’ values. The NN system compares these quantities with the
desired amount. The valves’ control system and the neural network controller can
say to operate simultaneously (depending on the selection of the system sampling
frequency) during the continuous closed loop cycles. The speed of the feedback
control system is set faster by the sampling period of the system. In doing so, we
also had to reduce the delay feedback output of the NN.

9.8.2 Process Verification

We collected eight input/output sets of random samples for the 3-1-3 neural net-
work configuration. These samples were collected randomly from the different
values to the input neurons’ layer [1 0.5 0.2;0.5 0.5 0.5;0.5 0.5 0.2;0.5 0.5 1;0.5
0.37 0.5;0.5 0.59 0.5;0.33 0.5 0.5;1 0.5 0.5] for matching to the same desired output
[1 0 1]. With the desired output fixed, the neural network backpropagated itself.
Each set of its output parameters stabilized itself automatically after adaptation. We
record the weights for each different sets of the input/output. The output neuron
parameters were plotted in the graph, through the formula as follow:
!
X
3
Catalyst ¼ W54 fW4i Xig ð9:18Þ
i¼1
!
X
3
Temperature ¼ W64 fW4i Xig ð9:19Þ
i¼1
!
X
3
Reactance ¼ W74 fW4i Xig ð9:20Þ
i¼1

Note: W4i refers to the neurons’ connected weighing values between the input
layer and hidden layer; X1 is the catalyst input; X2 the reactance input; X3 defines
to the temperature input signal (where Xi references to each of its valve’s variable).
The catalyst output value may vary during tuning, to match the desired reference
output of the neural network. That goes along with the auto adjustment of the
temperature for the reaction. As can be seen in Fig. 9.13 from sample 7 to 8, the
temperature automatically reduces as we increase the catalyst input. That is auto
tuning of the system to maintain the chemical reaction of the mixture we set. The
two parameters go according to a scaling rather than its unit. As we know, the
136 9 Machine Learning

Fig. 9.13 NN output neurons Output Neurons' Layer

(with secondary axis) 0.05 0
-0.02
0.04 -0.04

Temperature
Catalyst
0.03 -0.06
-0.08
0.02 -0.1
-0.12
0.01
-0.14
0 -0.16
1 2 3 4 5 6 7 8
Random Samples
Catalyst Temperature

function of the catalyst is to accelerate the reaction of the chemical mixture. So,
more catalyst will lower the activation energy require for more chemical reaction to
occur. Less catalyst makes the reaction slower. Thus, we increase the temperature
of the reaction or mixture to boost back the chemical reaction in the plant, to
produce the desired chemical mixture. The amount of heat boosted is equivalent to
the catalyst reduction during valves’ auto tunings. With a static valve positioning
for the reactance, the catalyst valve and the coolant valve auto adjusts themselves to
balance the mixture reaction required in the plant. The valves’ adjustment affects
the rate of the reaction and the quality of the mixtures, within the safe range of the
chemical production. Our aims are to meet these two requirements in the system.
Alternatively, our reference output for the neural network is provided to match
these two criterions. An advantage of the neural network control system is that we
do not need to calculate the error relationship between the valves’ settings and their
corresponded quantities of its product parameters. The continuous auto tunings of
the NN and the valves will reduce and eliminate the error deviation of the plant
parameters. Moreover, the control station will calculate the matching between the
actual plant and the desired chemical to correct the target signals. Figure 9.14
shows both the catalyst (top) and the temperature (below) for a set of the target.
The NN output neurons are calculated using Eqs. (9.18)–(9.20), from the eight
different input samples, to produce the same output target. The centerline shows the

Fig. 9.14 Neural network

output layer
9.8 Chemical Plant NN Feedback Control System 137

calculated average of the two parameters. We estimate the three curves by poly-
nomial equations. With the given average and any of the parameter in any timeline,
we can calculate the other parameters. The parameter found will lie along the
polynomial of the parameter itself. In this case, we can calculate either of the
parameters offline, to set the predictor limit for the parameters. Thus, we can predict
the parameters’ timelines also.

9.8.3 Process Improvement

So far, we have used the binary neurons for the NN system. Each neuron’s output
value is adjusted away from its real value during computations, for the neurons’
output function. Hence, the system accuracy is affected. Previously, we have set the
output layer neurons to its binary estimation. We lost the precision of the target
setting also, due to the estimation to the target binary. We want to improve the
system precision and accuracy. So we will now set the target output to their exact
real values, with a precision of two decimal points. For each neuron, we will use
their activation (sigmoid) function values as its output. Besides, we also update the
threshold values using the update formula as shown.

h j ðt þ 1Þ ¼ h j ðtÞ þ Dh ð9:21Þ

Dh ¼ aðh j ðtÞh j ðt  1ÞÞ þ gd j ð9:22Þ

where g is the gain term; α is the momentum term; θ is the thresholds term of each
node.
[Recall: Eq. (9.13) where fj ðnetÞ ¼ 1 þ exp  1netj þ h j ]
ð ð ÞÞ
For example, we set our desired target quality as [0.93 0.35 0.71]. Upon ini-
tialization, the starting weights are similar to Table 9.1. The initial threshold values
biased at 0. We simulate for the first eight sampling inputs as in the program 9.1.
We train the system and record their weights and biased values, after meeting its
performance goal. Our mean square error set at exponential −24. According to the
Eqs. (9.18)–(9.20), we trace the parameters’ updated trends according to each of its
input samples. We verify the first three input sampling sets [0.15 0.47 0.17],
[0.42 0.23 0.74], [0.41 0.58 0.2] for the trend. For the first to the second input
sample, we can see the rise in the catalyst but fall in the reactance. Besides, the
temperature rises for the input samples. As the transitions for the input samples are
the valves’ parameters which take place at the site, it enables more of the chemical
reaction to occur at the site. The neural network system automatic adjust itself to
control the mixture to meet the target quality set. Thereby, causing the catalyst and
the reactance mixture to reduce together with the targeted temperature. As we can
see in Fig. 9.15, the catalyst drops in accordance to the reactance. So the mixture
desired outcome is maintained. The target temperature for the mixture also
138 9 Machine Learning

Fig. 9.15 Auto adjustment NN Layer Output For Real Neurons Target
trends of the NN system

Scaled by E-11
Catalyst
Temperature
Reactance

Samples

maintains throughout the mixing. For the 2nd to 3rd sample input, the reactance
increases and the temperature dropped. It results in the slowing down of the
chemical mixture. The resultant NN control reacts to boost back the chemical
reaction to match the targeted parameters and chemical quality. We can clearly
observe the rise in the parameters as in the figure.

9.9 Remote Operated Neural Network Control Plant

9.9.1 Field Instrumentations

The Remote Operations Controller (ROC800) [25] is ideal for applications in

remote monitoring and control measurements. The 32-bit microprocessor-based
controller is used widely, especially for PID flow control applications. The recorded
database allows for the storing of events and the triggering of alarm signals in the
plant. We can measure up to a maximum of 12 monitoring signals for the instru-
ment. The wireless network radio modem provides the means for transmitting
information for the ROC800. The features of the system allow it to operate with an
update rate from 1 s to 60 min. The range of operating distance is from 230 m to
1 km. With the attractive characteristics, we can configure the ROCLINK 800
software to monitor the plant parameters. We can customize to control the plant also
with neural network control capability (Fig. 9.16).
The ControlWave Micro is with the IEC 62591 module installed. Software in the
control station manipulates the received signals for transmission back to the field.
OpenBSI software can configure the ControlWave Micro. There are several choices
of radio transmissions for the field parameters. The ROC mentioned is for discrete
signals only. Besides, we can use the Rosemount 248 wireless temperature trans-
mitter and the Fisher 4320 wireless valve position monitor for the system.
Additional choice includes the wireless smart THUM adapter [27] to sense the
valves’ parameters. These devices are suitable for use with the IEC 62591
9.9 Remote Operated Neural Network Control Plant 139

ControlWare Micro ControlWare

with IEC62591 Designer
module Software
Wireless
Valve

Fig. 9.16 IEC 62591 interfacing network

(WirelessHART). However, network radio module (NRM) [26] provides

bi-directional transmission of measured and controlled signals. It allows for mon-
itoring and intervention within the field operations. We can connect up the 2.4 GHz
NRM for the purpose with an update rate of 0.05 Hz. The valve scaled position
feedback calculates as:

Scaled Position ¼ ½ðUpper Range Value  Lower Range ValueÞ

ð9:23Þ
 Position % þ Lower Range Value

The valve sensor can also provide valve on/off control for overflow prevention in
the plant system. Moreover, it can be used to open/close the coolant valve for the
heat exchanger when the temperature reaches the target limit. The Westlock valve
ranges a distance of 110° positioning turns during on/off. The linear or rotary valve
reports every 1.5° turns. However, the transmission range is limited to only 100 m
with a router.

9.9.2 Scaling and Conversions

The scaled value Eq. (9.23) from the valve will go through a converter unit to
allocate the range into suitable values for the input neurons. Before that, the scaled
value had to be in units of angular degrees. The scaled value is equivalent to the
valve scaled position. The neuron range of 0–0.73 derives from the valve operating
range of 0–110° turns. The converted value lies within the range of 0–1 for the
neuron inputs.

Converter value ¼ ð½Scaled Value  Lower range value=1:5Þ  0:01 ð9:24Þ

So the valve sensors’ signals go through a conversion for matching the NN input
range. For the output layer neurons’ signals, we had to convert back the values to
the scaled values using Eq. (9.24), if the valves do not read the actual neurons’
values. We can refine the system to produce a more sensitive output. To do that, we
just change the output functions of the neurons to real values instead of binary
140 9 Machine Learning

values. We can use the ‘logsig’ logarithm sigmoid function so that the output range
stays between ‘0’ and ‘1’. All these can be done at the control station for the input
and output of the neural network system before transmissions.

9.9.3 Control Valves

The field electronic valve automation may consist of a single valve automation unit
or with another valve adapter for transmitting readings back to the control station.
A controller situates inside the electronic valve (see Fig. 9.18). It activates the PID
input, which receives the signal through an antenna. The actuators then manipulate
the valves from the PID signals. Altogether, at least, two throttle valve automation
sets are required for the system. Either the on/off valve or the throttle valve are used
for the heat exchanger to control the temperature of the mixture. Additional auto-
matic wireless on/off control valve can be used for the system also. It is for the
purpose of controlling the upper and lower limit of the chemical tank. The auto-
mated on/off valves for the mixer tank control the liquid overflow. We can use the
digital set-point to trigger the overflow alarm as well as to control the automated
on/off valve. A 4–20 mA level transmitter can be used to detect the level in the
tank. We can align the signal and scaled it to the 0–0.73 range for the digital valve.
Such that when the level is low at 0 mA, the valve will receive a 0.73 digital signal
to open the valve fully. If the level reaches the full 20 mA signal, it will activate the
valve with a zero signal to close it (Fig. 9.17).

9.9.4 Wireless Transmissions

The NN control system can be programmed to perform the function. We can insert
a delay sampler between the input neurons and the input weights of the neural
network. We program the system to detect the input signals from the valves at every
20 s interval. It is the same as delaying the system for every 20 s. Another way is to

Fig. 9.17 On/off valve control

9.9 Remote Operated Neural Network Control Plant 141

antenna controller

actuator
valve

Fig. 9.18 Wireless valve automation [32]. Courtesy of Emerson Process Management

Fig. 9.19 Automatic neural network control system

set the sampling period for the wireless transmitting rate to 0.05 Hz. The NN will
collect the input values to tune it to its desired output. The zero deviations in the
output signals halt the NN from further training. Then, each output parameters are
transmitted wirelessly to the valves. We set a transmitting delay at the output of the
NN system to the valves. This delay can be set to 10 s or so, as long as it lies within
the system sampling period (set at 20 s). The delay time allows for the automated
valves’ adjustment turns, as well as the valve sensor feedback and wireless trans-
mitting times. Moreover, it also caters for the neural network training and adap-
tation time.
The output sampler turns on again, and the valves accept new input values every
20 s. The closed loop system controls the valves for the desired mixture chemical
with controlled temperature limit (Fig. 9.19).
German engineering ‘JULABO’ [37], offers the state-of-the-art temperature
control technology for a wide range of temperature measurements from −95 °C to
over 200 °C. Several challenging features suitable for lab-based and integrated
outdoor environment, especially for the chemical reaction. Features such as fast
heat-up and cool down times for low power electronic controlled smart pumps
provides a more reliable control technology for heat exchange units. Moreover, the
142 9 Machine Learning

advanced control system operates without changing the bath fluids. Wireless tem-
perature control is available for the chemical process control system. We can easily
integrate the system for process optimization.

9.10 Valves and Chemical Plant Tunings

9.10.1 Desired Chemical Mixture, Samples and NN Data

The reference input to the system is the desired output of the neural network. We
conduct laboratory experiment for the quantity of the reactance [34] and catalyst
mixture allowed. The mixture is tested to match the quality requirement. We record
the required percentages of the mixture to produce the least heat temperature.
A proportional amount of the mixture is carried out in the actual plant with the
temperature monitored. We then collect the desired sample of the catalyst, reactance
and heat requirements without the NN system involved. Besides, we also collect
several input/output desired sets of the three valves’ settings as input parameters for
training the neural network. The valves’ settings and its parameters are the same.
Their ranges are between 0 and 1. The NN system is trained off-line to tune the
adapted weights and biased parameters to their stable sets of values. Finally, we are
ready to apply these NN parameters to the on-line system. In this way, the online
training time can be reduced.

9.10.2 Chemical and Valves Calibration

The control station sets the matching and desire three valves’ positionings with their
parameters (catalyst, reactance and temperature) for the chemical mixture. The flow
and level transmitters of the two small tanks transmit their signals back to the
station and compare the quality of the mixing chemical with the proportional
chemical property, collected from the lab samples. Besides, the temperature signal
at the mixture tank also transmits back to the control station. The quality relates to
the quantities of the reactance and catalyst mixed at the desired temperature. We
first calibrate the temperature of the mixer tank with the heat exchange control
valve, with reference to the catalyst and reactance. The flow and level of each tank
determine the quantities of its chemical. Once the quality of the mixture matches at
the desired temperature, the ideal valves positionings for the three control param-
eters will be set. Calibration can be at the control station, or at the field valve
automated system or both. So we control the amount of catalyst, reactance and heat
produced in the chemical mixing plant. The neural network system controls the
three control valves as shown in Fig. 9.20.
9.10 Valves and Chemical Plant Tunings 143

Fig. 9.20 Chemical plant valves and sensors

9.10.3 Trial Test in Actual Plant

The offline calculated NN parameters deploy to the real-time system. The trial test is
conducted at the actual plant to get the correct output that we want for the complete
system. We can adjust any initial settings of the three valves’ parameters for the
start. For every 20 s sampling delay, the NN receives three input values from the
valve sensors to calculate at least an iteration in the NN. Through the NN training,
the system can compute the correct output at the output neurons. These outputs
transmit through the delay, and then to the site control valves wirelessly. The valves
are adjusted automatically to produce the desired mixture within the controlled
temperature range. We verify the actual property of the mixture at the plant output.
The trial test completes once the quality satisfies.
There might be a problem that we are unable to boost the temperature up or cool
it down for the mixer tank. For example, when the catalyst input drops, the heat
might not be raised high enough to boost back the chemical reaction to take place.
Remember that we only use a coolant heat exchanger unit in the plant. Hence, our
remedy is to reinforce another heat exchanger to the system. The second exchanger
unit is to supply the heat source to raise the temperature enough for the chemical
reactions. So this heat exchanger is sourced with hot water instead of cooling water.
The control command for the valve of the second heat exchange unit (heat source)
will be opposite as the first (coolant source) unit. We define the valve manipulation
as:

CCE ¼ 1CHE ð9:25Þ

where CCE is the coolant source, and CHE is the heat source, they both control the
valve position variables of the heat exchanger units. Furthermore, we may use four
144 9 Machine Learning

Fig. 9.21 Two heat exchanger units

Fig. 9.22 Control valves operating ranges

input/output neuron signals for the neural network system to control the chemical
plant. The additional input/output neuron variable references to the hot water
exchanger control valve. We can then fine tune the internal neuron layers if it is
better than the 4-1-4 NN configuration we proposed. The control valves operating
ranges for the reactance is set similarly as the catalyst valve as shown in Fig. 9.22.
Whereas, we set the two heat exchanger valves in the opposite direction as
according to the definition Eq. (9.25). The cooling valve will not set to fully close
while the heating valve will not be set to fully open. These settings can lower the
presence of heat in the reactance at the initial stage (Fig. 9.21).

9.11 Computerized Neural Network Control System

9.11.1 NN Real Time Control Plant

Finally, we can implement the neural network to the real-time control system.
Training and adaptation are necessary on the online system to eliminate any
transmission network, instrumentations software and sensors errors. We can fit in
9.11 Computerized Neural Network Control System 145

the found NN parameters previously to reduce the training time. The NN param-
eters are automatically trained online again to adapt to the desired output of the
three parameters. The NN outputs are found and transmitted to tune each of the
corresponding valves. The continuous wireless closed loop network calibrates itself
to maintain the valves’ positions for the desired chemical output. In the system, we
can be aware of the on-site valves or instrumentation hardware failures if we detect
any changes in deviation in the actual chemical output with the set target (i.e. we
check for the matchings between the target quality (set valves) and the actual
chemical output quality, from the computerized advisor). Another advantage is that
we can change the desired output online by changing the reference signals to the
output neurons. The system will calibrate itself to its new weights and biased
values. Thus, the control valves changed according to the new values.

9.11.2 Neural Network Control Valves

For the program, we input the chosen catalyst value for the plant. So the selected
catalyst valve position is referenced or proportional to the catalyst variable. Unlike
the catalyst and the reactance, the temperature is inversely proportional to the heat
exchange valve position. Thus, we inverted and scaled the temperature output
neuron value to control its valve position. We can control the heat exchange valve
to perform the same way as the catalyst valve position variable. First, we need to
calibrate the temperature according to the valve’s variable. We calibrate the tem-
perature of the desired chemical tank output, with reference to the heat exchange
valve position. For example, a valve range from 0 to 0.73 is calibrated to the
temperature from 0 to 73 °C. We can adjust the degree of warm water in the heat
exchanger unit. Scaling is not necessary if the neurons’ value matches the valves’
range. The converted valve’s variable ranges from 0 to 0.73, for the heat valve,
disregards to the actual temperature readout in the mixer tank. The computer station
does the conversions. For the coolant valve, we convert the variable for the valve
positioning, by using the formula Eq. (9.24), with reference to Fig. 9.22. The heat
exchange valve position is transmitted back to the input neuron with the same
conversion and scaling as at the temperature output neuron. We maintain the
reactance input set valve position at a level of percentage tolerance (at ± desired
valve setting). Similarly, as the catalyst parameter, if we convert the output neurons
of the reactance to the control valve, it has to be inverted back from the control
valve variables to the input receiving neuron. Remember that we set and convert all
the three parameters of the valve’s positioning values to the input neurons. The
neurons accept values from 0 to 1 for computation. Figure 9.24 shows the training
convergence for the 3-1-3 neural network configuration. After training, we simulate
the result to a random set of input for the three neurons. The simulation converges
146 9 Machine Learning

Fig. 9.23 Neural network From Valves To Valves

I/O block

Fig. 9.24 NN training result Performance is 1.77237 e-025, Goal is 1e-024

Training-Blue Goal-Black

6 Epochs

Fig. 9.25 Simulated result Output Layer Neurons

Output Y

Input Sample
9.11 Computerized Neural Network Control System 147

to the target output after six epochs. We record the training weights as shown. The
output Y matches our desired values [0.93 0.35 0.71]. Figure 9.25 illustrates the
output result. However, we can increase the numbers of target tracked at the same
time by increasing the hidden neurons in the program (Fig. 9.23).

--------------------------------------------------
Program 9.1: Neurons Simulated Valves
--------------------------------------------------
real= [0.15 0.47 0.17; 0.42 0.23 0.74; 0.41 0.58 0.2; 0.09 0.51 1; 1 0.48 0.39; 1 0.37
1; 1 0.53 0.2; 0.82 0.56 1;0.17 0.5 0.43];
E=transpose(real)
p = [1 0.5 0.2]; !input captured or feed in from field database
P=transpose(p); !correct
t = [0.93 0.35 0.71; 0.93 0.35 0.71; 0.93 0.35 0.71; 0.93 0.35 0.71; 0.93 0.35 0.71;
0.93 0.35 0.71;0.93 0.35 0.71;0.93 0.35 0.71;0.93 0.35 0.71];
T=transpose(t); !the most only 2 different set binary targets can be tracked accurately
net = newff([0 1;0 1;0 1],[1 3],{'logsig' 'logsig'});
net.trainParam.show = 50; !only 1set of real target can be tracked accurately
net.trainParam.lr = 0.05;
net.trainParam.epochs = 38000;
net.trainParam.goal = 1e-24;
net.trainParam.gradient = 1e-15;
net.iw{1,1}=[0.3 0.5 0.1] ; !initial data
net.lw{2,1}=[0.1; 1; 0]; !Initialised data
net.b{1,1}=[0]; ! biased at ‘0’ to reach best performance goal
net.b{2,1}=[0; 0; 0]
[net,tr]=train(net,E,T);
iw=net.iw{1,1}
lw=net.lw{2,1}
bias=net.b{1,1}
bias2=net.b{2,1}
Y = sim(net,P)
Tar=[0.93 0.35 0.71];
Target=transpose(Tar)
input=[1];
plot(input,Target,'r+',input,Y,'ko')
hold; xlabel('Input Sample');
ylabel('Output Y'); title('Output Layer Neurons'); hold;
----------------------------------------------------------------------------------
Output Results : Output Y matches the target, with a given input P
----------------------------------------------------------------------------------
TRAINLM
Epoch 0/38000, MSE 0.101745/1e-024, Gradient 1.43031/1e-010
Epoch 6/38000, MSE 1.77237e-025/1e-024, Gradient 1.78259e-014/1e-010
Performance goal met.
148 9 Machine Learning

iw = 1.0e – 009 * – 0.0013

1.0e – 009 *0.2724
1.0e – 009 * 0.0274

lw = 0.9733
0.3275
0.2445

bias = 0.0014

bias2 = 2.0997
– 0.7829
0.7731

Y = 0.9300
0.3500
0.7100

Target = 0.9300
0.3500
0.7100

9.11.3 Intelligent Advisor

We can establish on-line purity assessment and track the chemical reactions of the
plant. Equipment from ‘Magritek’ can be used to monitor the state of the plant.
Spinsolve software can be customized to track the plant reactions. Moreover, we
can program a neural network system to advise on the differences between the
actual chemical plant and the desired target quality (see reference [36] DeltaV
system). The plant feedback variables from the automation, system or the chemical
software are stored and retrieved by the computerised advisor. We can return
position variables from the field valves into the computer neural network advisor by
using ODBC. So, all the field instrumentation information, including the valve
variables are updated continuously and stored in the database for retrieval.
9.11 Computerized Neural Network Control System 149

Fig. 9.26 Quality matching Output Target Quality

advisor

CIRCLES–Feedback Input Neurons (Quality)

ASTERIKS –Trained Target Samples

The NN advisor performs training for eight sampling targets for the chemical
desired qualities. The neural network for the advisor uses 3-5-1 NN configuration.
Each desired target shows in asterisk in the figure. For example, we select a target
quality (target sample 7) at 0.38. So we feed in reference valves input [1 0.83 0.2]
as the three valves variables to produce the target quality. The NN plant control
system (P9.1) operates, and control its valves to their positions in the field. If the
catalyst valve is stucked at position 0.89 (should be 1), the quality advisor will
receive the mismatch target at 0.32 (target sample 3). Thus, the incorrect or not
matching target lead us to troubleshoot the plant or valve error at the field site.
Hence, the intelligent advisor system determines the site hardware failure to activate
an alarm. Following up, we implement the fault correction at the field site
(Fig. 9.26).
150 9 Machine Learning

------------------------------------------------------------
Program 9.2: Computerized NN Quality Advisor
------------------------------------------------------------

real= [0.15 0.47 0.17; 0.42 0.23 0.74; 0.41 0.58 0.2; 0.09 0.51 1; 1 0.48 0.39; 1 0.37
1; 1 0.83 0.2; 0.82 0.56 1;0.17 0.5 0.43];
E=transpose(real)
t = [0.8204 ;0.5678 ;0.32 ;0.58 ;0.679 ;0.45 ;0.38 ;0.12 ; 0.42];
T=transpose(t); !Target qualities trained
net = newff([0 1;0 1;0 1],[5 1],{'logsig' 'logsig'});
net.trainParam.show = 50; !3-5-1 configuration to train more targets
net.trainParam.lr = 0.05; !at a time.
net.trainParam.epochs = 38000;
net.trainParam.goal = 1e-24;
net.trainParam.gradient = 1e-15;
net.iw{1,1}=[0.3 0.5 0.1;0.3 0.5 0.1;0.3 0.5 0.1;0.3 0.5 0.1;0.3 0.5 0.1]
net.lw{2,1}=[0 0 0 0 0]; !initialisztions
net.b{1,1}=[0; 0; 0; 0; 0]
net.b{2,1}=[0]
W1 = net.iw{1,1}
W2 = net.lw{2,1}
b1= net.b{1,1}
b2= net.b{2,1}
[net,tr]=train(net,E,T); !trainings
iw=net.iw{1,1}
lw=net.lw{2,1}
bias=net.b{1,1}
bias2=net.b{2,1}
Y1 = sim(net,E)
T
conne=database('db1','',''); ! on-line updated database stored in db1
bo = exec(conne,'select all Field1 from Table1');
sor=fetch(bo,3);
a=sor.Data{1,1};
b=sor.Data{2,1};
c=sor.Data{3,1};
inp=[a;b;c]
! inp = [0.89; 0.83 ;0.2]; incoming 3 valves feedback retrieved from db1
Y = sim(net,inp); !simulated quality feedback from the field
iw=net.iw{1,1}
lw=net.lw{2,1}
bias=net.b{1,1}
bias2=net.b{2,1}
input=[1 2 3 4 5 6 7 8 9]; figure(2);
plot(input,Y1,'k*')
hold; xlabel('ASTERIKS - Trained Target Sample');
9.11 Computerized Neural Network Control System 151

ylabel('CIRCLES - Feedback Input Neurons'); title('Output Target Quality');

plot(input,Y,'bo');
hold;
gtext('The target quality is set')
gtext('at 0.38, at the 7th sample')
Chapter 10
Computer Vision

Computer Vision is popular in the present automation, and control applications in

the modern industries. Manufacturing and assembly lines made use of the tech-
nology to track and scan for product defects in the production lines. Cameras are
fixed at the assembly lines to scan for bar code presence and other product
parameters in the processed images.

10.1 Image Thresholding

In computer vision, we often encounter problems of errors when all the pixels has
been thresholded. We often encounter two types of errors as a result of the extremes
of the two final level classifications as either foreground or background.
1. We are not able to catch all the included pixels in the group.
2. Some pixels caught should not be in the group.
So the choice of thresholding comes in, to balance these two types of error. The
threshold level is between 0 and 255 with reference to the black or the white
background respectively. We implement a two steps thinning algorithm without
violating the following constraints.
1. Does not remove end points.
2. Does not break connectivity.
3. Does not cause excessive erosion of the region.

© Springer Science+Business Media Singapore 2016 153

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0_10
154 10 Computer Vision

10.2 Zhang-Suen Thinning Algorithm

The mark-and-delete thinning algorithm has a template. The template is of size

3 × 3. It moves to overlap the image to determine the center pixel. It is an iterative
algorithm and continues until no more black seeds removed. We used two
sub-iterations in the algorithm.
First-step sub-iteration:
1. Connectivity (number of background-to-foreground, or ‘255’ to ‘0’ in this case)
in template boundary pixel is one.
2. The foreground or ‘0’ neighbouring pixels of the selected center template is
within 2–6.
3. Either one of the pixels P(y, x − 1) P(y + 1, x) or P(y, x + 1) must be a white
background.
4. Either one of the pixels P(y + 1, x) P(y, x + 1) or P(y − 1, x) must be a white
(255) background.
Second-step sub-iteration:
First two procedures are the same as the first-step sub-iteration. The third and fourth
steps are as follow:
3. Either one of the pixels P(y − 1, x) P(y, x + 1) or P(y, x − 1) must be a white
background or value of 255.
4. Either one of the pixels P(y − 1, x) P(y + 1, x) or P(y, x − 1) must be a white
background.
We flagged the seeds or pixels that satisfy the above conditions for changing.
The flagged pixels changed after each step. The two-step process repeats until we
can no longer change any more black pixels.

10.3 Brief Descriptions of the Program Algorithms

The program consists of two separate programs. Mainly the binary .m file and the
thinning .m file. The binary program includes the ‘Binary’ function and the ‘Binarise’
sub-function. The main function process the seven different kinds of the images to
read in and the different threshold level. It will call the binaries sub-function to do the
binary processing. The outcome of this program will create seven images stored into
different filenames each. The outcome is the binary images produced.
The second program ‘thinning .m’ which has the thinning function to do the
thinning process. Besides, it also has six different sub-function to support the main
program function. They are mainly the function: initialize, change, searchblack,
transition, steponecd and steptwocd. The initialize function is to initialize the
replacement data “AData” and the stored data “SData”. After which the change
function is to convert the black seeds or pixels into white pixels in the image. It is
10.3 Brief Descriptions of the Program Algorithms 155

the thinning process. The ‘searchblack’ function is to do the neighbouring search

for the black seeds on the required centered pixel. It searches for the eight sur-
rounding pixels to find the black seeds. If the black pixels neighbour is between 2
and 6, then it activates a one for recognition in the main process of the program.
The transition function finds a zero to one transition of the neighbouring pixels.
There must be one transition only, to initiate for processing. The next function is to
test for step 1c and 1d of the thinning algorithm. The last function ‘steponecd’ is to
condition for the step 2, c and d conditions to meet its requirements. For the last two
functions, a one will be activated if they meet the requirements. So altogether, the
four ones will go through a judging process, whereby only three of the ones will be
combined to check for pixels replacement into white seeds. Each of Zhang Suen’s
two step algorithm will test for three criteria, which will result in producing a 3
when we meet the four-condition of each step. This number of three will activate for
conversion of the black seeds into white seeds.
In the Matlab thinning main program provided, the four conditions in each of the
two steps are developed such that it only consists of three criteria in judging for
removal. We will test each of the first and second conditions by going through a
function subroutine. While we combine the third and fourth conditions to form
criteria in conditioning for pixels removal. We apply the same to the third and
fourth conditions in the second step, where only a function subroutine is used to test
for the two conditions. Therefore, only 3 test functions criteria are needed to be
satisfied in each step for pixel removal. The thinning algorithm consists of four
conditions of which we mix the last two conditions to form criteria. The main
algorithm for thinning is such that it loops through the two steps.
The program started off by reading into the program the image to thin. We take
into account the size of the picture. As the width and the height, of the picture
corresponds to the x, and y, which is the columns and rows of the image pixels
concerned. Next, the main program of the thinning function initialized the matrices
to be used. The thinning routine runs through pixel by pixel to search for the black
seeds or pixels. Once we find it, it will be tested in the 1st step of the thinning
algorithm. If it meets the three criteria of the first step, the test pixel or black seed
will be flagged for replacement.
If not, the black seeds will be stored in another matrix location for testing during
the second step of the algorithm. When all the pixels of the image are scan, the
program will go to the change function to get the flagged seeds to be replaced by
the white seeds to become the white background. Next, the remaining stored black
pixels will be tested by the second step algorithm for replacement with the white
seeds. It will go to the same procedure for pixels replacement once we flagged it.
Otherwise, the remaining black seeds will be stored again for testing in the next
round or iteration. The program repeats the two-step algorithm after the first two
steps are over. It will search for the remaining stored black seeds for replacement
with the white seeds. If no black pixels are to flag for replacement, then the program
comes to an end. The final output becomes the thinned or skeleton (media axis) of
the original image. We have to name the final image to store the final result. Zhang
Suen’s thinning algorithm thus provides the same outcome as the media axis
156 10 Computer Vision

transformation technique in gray level. It is most suitable for using as a skeleton

transforming method in binary level.

------------------------------------------
Program 10.1 Binarise Algorithm
------------------------------------------
%Binarise Original Images
function Binary
global map imga h CData Y height width t Y1 image
image=’img1.bmp’ %define image
[imga,map]=imread(image); %input image
height=336; %image height
width=339; %image width
t=150; %threshold level
figure(1) %open figure to view ready image
Binarise %apply thresholding to view the character clearly
imwrite(Y1,map,’bimg1.bmp’); %save image as a filename
image=’img3.bmp’ %an image is chosen
[imga,map]=imread(image);
height=349;
width=357;
t=70; %threshold level is input defined (changeable)
figure(3)
Binarise
imwrite(Y1,map,’bimg3.bmp’); %save as a different filename
image=’img4.bmp’
[imga,map]=imread(image);
height=746;
width=668;
t=70;
figure(4)
Binarise
imwrite(Y1,map,’bimg4.bmp’);
image=’img5.bmp’
[imga,map]=imread(image);
height=535;
width=576;
t=200;
figure(5)
Binarise
imwrite(Y1,map,’bimg5.bmp’);
image=’img6.bmp’
[imga,map]=imread(image);
height=436;
width=469;
t=70;
figure(6)
Binarise
imwrite(Y1,map,’bimg6.bmp’);
function Binarise %Binary function
global imga map Y CData h height width t Y1
10.3 Brief Descriptions of the Program Algorithms 157

h=image(imga);
colormap(map);
Y=get(h,’CData’); %grap image data (CData)
CData(1:height,1:width)=Y
for i=1:height %thresholding
for j=1:width
if CData(i,j)>=t
CData(i,j)=255;
elseif CData(i,j)<t
CData(i,j)=0;
end
end
end
Y1=CData(1:height,1:width) %Binarised matrix value
imwrite(Y1,map,’bimg.bmp’); %store temporary image in file
[Y1,map]=imread(‘bimg.bmp’); %read binarised image
imshow bimg.bmp; %show image

----------------------------------------==-
Program 10.2 Thinning Algorithm
--------------------------------------==---
function Thinning %thinning algorithm
global width height x y CData Yes AData seed SData
height=336; %image height depending on image input
width=339; %image width depending on image input
Yes=0; %initialisation
seed=0;
convert=0; %initialise convert
initialise; %initialise matrix
[Y1, map]=imread(‘bing1.bmp’); %changeable input image
h=image(Y1);
colormap(map);
Y=get(h,’CData’);
CData(1:height,1:width)=Y;
for y=1:height %Perform Step1
for x=1:width
if CData(y,x)==0 %if black seed found
searchblack; %do the 3 criterior of step 1
out1=Yes; %or the 4 conditions of step 1
transition;
out2=Yes;
steponecd;
out3=Yes;
out1+out2+out3;
if out1+out2+out3==3 %if satisfy the 3 criterior to convert
convert=1; %activate convert
AData(y,x)=CData(y,x); %flagged for conversion
else
SData(y,x)=CData(y,x); %store black seed for next round thinning decision
158 10 Computer Vision

seed=seed+1; %indicate seed for thinning decision

end
end
end
end
change; %convert black seed into white
for y=1:height %Perform Step 2
for x=1:width
if (SData(y,x)==CData(y,x)) %if the stored black seed is found
searchblack; % do the 3 criterior of step 2 again
out1=Yes; % or the 4 conditions of step 2 again
transition;
out2=Yes;
steptwocd;
out3=Yes;
out1+out2+out3;
if out1+out2+out3==3 % judge the 3 criterior
convert=1; %activate convert
AData(y,x)=CData(y,x);
seed=seed-1; %indicate seed left to be thinned
else
SData(y,x)=CData(y,x);%else store the remaining black seed for next round
end
end
end
end
change;
while convert>0 %loop until no black seed is to be removed
convert=0; %reinitialise convert
for y=1:height %Perform Step 1
for x=1:width
if SData(y,x)==CData(y,x) %if black seed found
searchblack;
out1=Yes;
transition;
out2=Yes;
steponecd;
out3=Yes;
out1+out2+out3;
if out1+out2+out3==3
convert=1; %activate convert
AData(y,x)=CData(y,x);
seed=seed-1;
else
SData(y,x)=CData(y,x);
end
end
end
end
change;
for y=1:height %Perform Step 2
10.3 Brief Descriptions of the Program Algorithms 159

for x=1:width
if SData(y,x)==CData(y,x) % if black seed found
searchblack;
out1=Yes;
transition;
out2=Yes;
steptwocd;
out3=Yes;
out1+out2+out3;
if out1+out2+out3==3
convert=1; %activate convert
AData(y,x)=CData(y,x);
seed=seed-1;
else
SData(y,x)=CData(y,x);
convert=0; %check for no convert, may be removed for total thinning
seed
end
end
end
end
change;
end
Y1=CData(1:height,1:width) %Binarise matrix value
imwrite(Y1,map,’timg11.bmp’); % store image into a file
[Y1,map]=imread(‘timg11.bmp’); %readin image to be displayed
figure(2);
imshow timg11.bmp; %display image

function initialise %initialise AData & SData

global x y height width AData SData
for y=1:height
for x=1:width
AData(y,x)=150; %initialise to intermediate value
SData(y,x)=150; %of between black and white
end
end

function change %black seed change into white

global CData AData x y height width
for y=1:height
for x=1:width
if AData(y,x)==CData(y,x)
CData(y,x)=255;
end
end
end

function searchblack %finding black neighbours

global CData x y Yes
b=0;
160 10 Computer Vision

Yes=0; b=b+1; end

if CData(y-1,x-1)==0 b;
b=b+1; end if (b>=2)
if CData(y+1,x+1)==0 if (b<=6)
b=b+1; end Yes=1; end
if CData(y+1,x)==0 end
b=b+1; end
if CData(y-1,x)==0
b=b+1; end
if CData(y,x+1)==0
b=b+1; end
if CData(y,x-1)==0
b=b+1; end
if CData(y+1,x-1)==0
b=b+1; end
if CData(y-1,x+1)==0
function transition %zero to one transition
global Yes CData x y
rt=0;
if (CData(y,x-1)>0) if (CData(y-1,x+1)>0)
if(CData(y+1,x-1)<255) if (CData(y-1,x)<255)
rt=rt+1; end rt=rt+1; end
end end
if (CData(y+1,x-1)>0) if CData(y-1,x)>0
if (CData(y+1,x)<255) if CData(y-1,x-1)<255
rt=rt+1; end rt=rt+1; end
end end
if (CData(y+1,x)>0) if CData(y-1,x-1)>0
if (CData(y+1,x+1)<255) if CData(y,x-1)<255
rt=rt+1; end rt=rt+1; end
end end
if (CData(y+1,x+1)>0) rt;
if (CData(y,x+1)<255) if rt==1
rt=rt+1; end Yes=1;
end else
if (CData(y,x+1)>0) Yes=0;
if (CData(y-1,x+1)<255) end
rt=rt+1; end
end
function steponecd %Step 1c and 1d conditions
global Yes CData x y
Yes=0;
if (CData(y,x-1)+CData(y+1,x)+CData(y,x+1))>0
if(CData(y+1,x)+CData(y,x+1)+CData(y-1,x))>0
Yes=1;
end
end
function steptwocd %step 2c and 2d conditions
global CData x y Yes
Yes=0;
if (CData(y,x-1)+CData(y-1,x)+CData(y,x+1))>0
if (CData(y,x-1)+CData(y-1,x)+CData(y+1,x))>0
Yes=1;
end
end
10.4 Image Results 161

10.4 Image Results

In the first five sets of the character examples, pictures two, three, and five, give the
same amount of thresholding of about 70 into the image, while that of image 1,
thresholds at 150 and image four thresholds at a value of 200. For image one, when we

Original Image 1 Binarised T=150 Thinned

Binarised T=230 Thinned

Binarised T=70 Thinned

(Total white image) (Total white image)

Original Image 2 Binarised T=150 Thinned

Binarised T=200 Thinned

Binarised T=70 Thinned

Fig. 10.1 Chinese characters (images sets 1 and 2). Note T represents threshold level within
0–255
162 10 Computer Vision

Original Image 3 Binarised T=150 Thinned

Binarised T=200 Thinned

Binarised T=70 Thinned

Thinned at T=10 (break connectivity)

Original Image 4 Binarised T=230 Thinned

Binarised T=200 Thinned

Thinned at T=50

Original Image 5 Binarised T=200

Binarised T=70 Thinned

Binarised T=10 Thinned

Fig. 10.2 More Chinese characters (images sets 3–5)

10.4 Image Results 163

input a binarized level of 150 will perform the thinning well. If a value of 230 is input,
the resulting character after thinning will still be fined. But if we apply a threshold
level of 70, it will result in a blank image after binarized, so is after thinning.
Image 2 is binarized at a value of 70. We can binarize the image at a level of 10
also. The outcome after thinning is as good. However, if we use a value of 200 or
150 as the threshold level, the binarized image will produce some black patches.
And if thinned, the character will not be a skeleton. It will still be fat with only
minor trimmings of the character.
The black and dark original image 3, if binarized at 200, will cause the creation
of the black border boundary, which is far away from the character, to appear in the
image after thinning. It is the result of the excessive level of the threshold value
applied. However, when thresholded at 150, the final thinning result will still create
a slightly fatter image character. It is not the full thinning operation, and we cannot
produce the skeleton outcome. On the other hand, when binarized at a value of
about 10, will result in the breaking connectivity of the character when after
thinning. Image 4 has a faint image, so a higher value of the thresholding is needed.
Image 4 erodes when thresholded at a level of 50. The final result of the eroded
character after thinning will be a blank white image without any character in view.
If binarized at a value of 230, or slightly over thresholded, the outcome will be a
slightly fatter skeleton, which will still look like a character, but only never
skeletoned. Moreover, the borderline of the image will appear.
In image 5, the average threshold is at 70–10. It is because it has a dark original
image background, so lesser thresholding level is required. However, if we bina-
rized at a very large value of 200, the character itself will be over darken and tends
to become overconnected as if black ink leaks out from the character. Besides,
black dots started to appear in the image from nowhere. The character will worsen
when we thinned it. It will become a dark patch of an unrecognised character
(Figs. 10.1, 10.2).

Original Image 6 Binarised T=200 Thinned

Binarised T=70 Thinned

Fig. 10.3 Hao Chinese character (image set 6)

164 10 Computer Vision

Original Image 7 (8-bit depth bitmap)

Binarised T=70 Thinned at T=70

Binarised T=200 Thinned at T=200

Fig. 10.4 Scanned original character (image set 7)

So thresholding adjustments are for improving the appearance and visibility of

the characters at the binary level, which will in turn affect the thinning result of the
characters. Let us look at the image set 6. The set of image thinned lesser at a lower
threshold of 70 than at threshold level of 200. We can see that it has the reverse
thinning effect from the rest of the image sets. Only this original image set is
exceptional. Moreover, it is noted that the scanned image (image set 7) after
binarization, cannot produce the proper characters for thinning. It is because the
10.4 Image Results 165

given image is only an 8-bit colour depth image. It represents only one (8-bit) out of
the three primary colour (24-bit) components. Zhang Suen’s method of thinning
provides an erosion of the original image into a skeleton. Likewise, it can also just
thinned a bit, which results in the thick character still remainings. The thresholding
level affects the thinnings of the characters. If the threshold level is too high towards
white, applying thinning algorithm will convert more to the black side which will
eventually darken the image. However, if the threshold level is small towards the
value of black (zero), applying thinning will convert lesser of the value to the black
region thereby causing connectivity brokage of the character. It, therefore, results in
the eroding of the character from the original image. That is what happens to the
final results to some of the characters applied (Figs. 10.3, 10.4).
Appendix A
MC68HC11 Registers

def SPDR 102A u8

def PORTA 1000 u8 def BAUD 102B u8
def DDRA 1001 u8 # 68HC11F1... def SCCR1 102C u8
def PIOC 1002 u8 def SCCR2 102D u8
def PORTG 1002 u8 # 68HC11F1... def SCSR 102E u8
def DDRG 1003 u8 # 68HC11F1... def SCDR 102F u8
def PORTC 1003 u8 def ADCTL 1030 u8
def PORTB 1004 u8 def ADR1 1031 u8
def PORTCL 1005 u8 def ADR2 1032 u8
def PORTF 1005 u8 # 68HC11F1... def ADR3 1033 u8
def PORTCF1 1006 u8 # 68HC11F1... def ADR4 1034 u8
def DDRC 1007 u8 def OPT2 1038 u8 # 68HC11F1...
def PORTD 1008 u8 def OPTION 1039 u8
def DDRD 1009 u8 def COPRST 103A u8
def PORTE 100A u8 def PPROG 103B u8
def CFORC 100B u8 def HPRIO 103C u8
def OC1M 100C u8 def INIT 103D u8
def OC1D 100D u8 def TEST1 103E u8
def TCNT 100E u16 def CONFIG 103F u8
def TIC1 1010 u16 def CSSTRH 105C u8 # 68HC11F1...
def TIC2 1012 u16 def CSCTL 105D u8 # 68HC11F1...
def TIC3 1014 u16 def CSGADR 105E u8 # 68HC11F1...
def TOC1 1016 u16 def CSGSIZ 105F u8 # 68HC11F1...
def TOC2 1018 u16 def SCONF 0420 u8 # X68C75
def TOC3 101A u16 def PORTBO 0410 u8 # X68C75
def TOC4 101C u16 def PORTBI 0430 u8 # X68C75
def TOC5 101E u16 def PORTCO 0408 u8 # X68C75
def TCTL1 1020 u8 def PORTCI 0428 u8 # X68C75
def TCTL2 1021 u8
def TMSK1 1022 u8
def TFLG1 1023 u8
def TMSK2 1024 u8
def TFLG2 1025 u8
def PACTL 1026 u8
def PACNT 1027 u8
def SPCR 1028 u8
def SPSR 1029 u8

© Springer Science+Business Media Singapore 2016 167

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0
168 Appendix A: MC68HC11 Registers

' startup.bas
ProgramPointer $8000
DataPointer $2000
StackPointer $7FFF

sect text
cli ; enable debugger
ldx #_data_s
bra _crt2
_crt1 clr 0,x ; clear data area
inx
_crt2 cpx #_data_e
bne _crt1
sect data
_data_s equ *

byte PORTA at $1000 ' mode single-chip

byte DDRA at $1001
byte PORTG at $1002
byte DDRG at $1003
byte PORTF at $1005
byte PORTD at $1008
byte DDRD at $1009
byte PORTE at $100A
int TCNT at $100E
byte TCNTL at $100F
byte TMSK2 at $1024
byte TFLG2 at $1025
byte PACTL at $1026
byte PACNT at $1027
byte BAUD at $102B
byte SCCR1 at $102C
byte SCCR2 at $102D
byte SCSR at $102E
byte SCDR at $102F
byte ADCTL at $1030
byte ADR at $1031
byte OPTIONS at $1039
byte PORTB at $1060
byte PORTC at $1061
byte PORTM at $1062
byte PORTN at $1063
Appendix B
MCU Port Testers

© Springer Science+Business Media Singapore 2016 169

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0
170 Appendix B: MCU Port Testers

porta=porta and $df

delay(10)
porta=porta or $ff
delay(10)
porta=porta and $bf
delay(10)

'PORTA TESTER

ProgramPointer $2000 porta=porta or $ff

DataPointer $0002 delay(10)
StackPointer $7FEF porta=porta and $7f
delay(10)
byte ddra at $1001 next q
byte porta at $1000 porta=porta or $ff
int q,s for s=1 to 10
ddra=$ff porta=porta xor $ff
porta=$ff delay(100)
ASM cli next s
do loop
q=0;s=0
for q=1 to 3 function delay(c)
porta =porta or $ff int i
delay(10) for c=c to 0 step -1
porta=porta and $fe for i= 0 to 1000
delay(10) next i
porta=porta or $ff next c
delay(10) end function
porta=porta and $fd
delay(10)
porta=porta or $ff
delay(10) ''PORTB TESTER
porta=porta and $fb
delay(10) 'byte ddrb at $1001
porta=porta or $ff byte porta at $1004
delay(10) int q,s
porta=porta and $f7 porta=$ff
delay(10) ASM cli
porta=porta or $ff do
delay(10) porta=$55
porta=porta and $ef delay(100)
delay(10) porta=$aa
porta=porta or $ff delay(100)
delay(10) loop
Appendix B: MCU Port Testers 171

function delay(c) 'PORTD TESTER

int i
for c=c to 0 step -1 byte portd at $1008
for i= 0 to 1000 byte ddrd at $1009
next i byte portg at $1002
next c byte ddrg at $1003
end function
ddrd=$3f
ddrg=$0c
portd=$f0
'PORTC TESTER ASM cli

int j,f
byte portg at $1002 do
byte ddrg at $1003 portg=$07
ddrg.1 = 1 delay(300)
byte portc at $1061 'no need portg=$0b
define ddrc delay(300)
'byte title()=" Microprocessor" portg=$ff
'byte portc at $1006 'single chip portd=$df
mode no need define ddrc delay(300)
for j=100 to 0 step -1 portd=$ef
for f= 0 to 1000 delay(300)
portg.1=0 portd=$f7
next f delay(300)
next j portd=$fb
ASM cli delay(300)
do portd=$ff
if portc.4=0 then portg.1=0 loop
if portc.4=1 then portg.1=1
portc.4=0 'useless command function delay(c)
loop int i
for c=c to 0 step -1
function delay(c) for i= 0 to 1000
int i next i
for c=c to 0 step -1 next c
for i= 0 to 1000 end function
next i
next c
end function
'PORTE TESTER

ProgramPointer $8000
DataPointer $0002
StackPointer $7FEF
172 Appendix B: MCU Port Testers

byte porte at $100A byte portg at $1002

byte ADCTL at $1030 byte ddrg at $1003
byte ADR at $1031 ddrg.1=1
byte OPTIONS at $1039 ASM cli
ADR=$00 do
OPTIONS=$90 'direct debug at $1063
delay(5) 'read(0)
ASM cli 'portn.7=0 'useless command
do analogout(4, 0xff)
analogin(2) 'BIT NO. 7 if portn.4=0 then portg.1=0
loop if portn.4=1 then portg.1=1
loop
function delay(c)
int i function analogout(ch, val)
for c=c to 0 step -1 ' 1:A, 2:B, 4:C
for i= 0 to 1000 portm=0x20 'cs=0
next i max512(ch)
next c max512(val)
end function portm=0x30 'cs=1
end function
function analogin(ch)
ADCTL = ch function max512(val)
do byte cnt
loop until ADCTL.2=1 for cnt=0 to 7
return ADR if val and 0x80 then
end function portm=0x22 'sdin=1
portm=0x23 'sclk=1
function read(value) else
int cha, old_cha portm=0x20 'sdin=0
if cha==old_cha then return 0 portm=0x21 'sclk=1
old_cha = cha end if
return cha portm=0x20 'sclk=0
end function val=val+val 'shift left
next cnt
end function

'PORTF TESTER function delay(c)

int i
ProgramPointer $8000 for c=c to 0 step -1
DataPointer $2000 for i=0 to 1000
StackPointer $7FEF next i
byte portf at $1005 next c
byte portc0 at $0408 end function
byte portn at $1063
byte portm at $1062
Appendix C
LCD References

See Tables A.1 and A.2.

Table A.1 Control and display command

© Springer Science+Business Media Singapore 2016 173

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0
174 Appendix C: LCD References

Table A.2 The command control codes

Command Binary He*
D7 D6 D5 D4 D3 D2 D1 D0
Clear display 0 0 0 0 0 0 0 1 01
Display and cursor home 0 0 0 0 0 0 1 x 02 or 03
Character entry mode 0 0 0 0 0 1 1/D S 04 to 07
Display on/off and cursor 0 0 0 0 1 0 U B 08 to 0F
Display/cursor shift 0 0 0 1 D/c R/L X x 10 to IF
Function set 0 0 1 8/4 2/1 10/7 X x 20 to 3F
Set CGRAM address 0 1 A A A A A A 40 to 7F
Set display address 1 A A A A A A A 80 to FF
l/D: 1 = increment*, 0 = decrement
S: 1 = Display shift on. 0 = Display shift off*
O; 1 = Display on, 0 = Display off*
U: 1 = Cursor underline on, 0 = Underline off*
B: 1 = Cursor blink on, 0 = Cursor blink off*
D/C: 1 = Display shift, 0 = Cursor move
R/L: 1 = Right shift, 0 = Left Shift
B/4: 1 = 8 bit interface*, 0 = 4 bit interface
2/1: 1 = 2 Iine mode, 0 = 1 line mode*
10/7: 1 = 5 × 10 dot format. 0 = 5 × 7 dot format*
x = Don’t care
* = Initialisation settings
Appendix C: LCD References 175
References

1. http://www.datasheet4u.com/datasheet/7/4/L/74LS08_FairchildSemiconductor.pdf.html
(2-input AND gate)
2. http://www.datasheet4u.com/datasheet/7/4/L/74LS32_FairchildSemiconductor.pdf.html
(2-input OR gate)
3. http://www.datasheet4u.com/datasheet/7/4/L/74LS04_FairchildSemiconductor.pdf.html (hex
inverting gate)
4. http://html.alldatasheet.com/html-pdf/171559/TI/74LS244/24/1/74LS244.html (Buffer or
Line driver)
5. http://www.ti.com/lit/ds/symlink/sn74ls74a.pdf (Dual D-Latch)
6. http://docs.google.com/viewer?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttp%2Fwww.datasheet.hk%2Fdownload_
online.php%3Fid%3D1027398%26pdfid%3D384EA12E7F41ADD153FD14E6B124409A
%26file%3D0021%5Csn74ls76_186750.pdf&embedded=true (Negative-edge triggered J-K
flip-flop)
7. http://www.alldatasheet.com/datasheet-pdf/pdf/22437/STMICROELECTRONICS/L298.html
(Motor driver)
8. http://www.pepperl-fuchs.com/global/en/classid_142.htm (proximity sensor)
9. http://ph.parker.com/sg/en/solenoid-valves (soleniod valve)
10. http://pdf.datasheetcatalog.com/datasheet/siemens/BUZ12.pdf (BUZ12)
11. http://www.datasheetarchive.com/dl/Datasheet-093/DSA0066289.pdf (IN4001 diode)
12. http://cache.freescale.com/files/microcontrollers/doc/data_sheet/M68
13. HC12B.pdf?pspll=1 (MCU 68HC12 family)
14. http://html.alldatasheet.com/html-pdf/50893/FAIRCHILD/7407/403/1/7407.html
(open-collector output hex buffer)
15. http://www.datasheetarchive.com/dlmain/Databooks-2/Book259-13.pdf (4N28 opto-coupler)
16. http://www.alldatasheet.com/datasheet-pdf/pdf/8979/NSC/LM555.html (555 timer)
17. http://www.ti.com/lit/ds/symlink/uln2003a.pdf(darlington transistor)
18. http://courses.cs.tau.ac.il/embedded/docs/LPC2148_Education_Board/KS0070B.pdf (LCD
dot matrix)
19. http://www.ee.nmt.edu/*rison/ee308_spr98/supp/feb_9/hc11_h.html (hc11.h file).
20. The Intel Microprocessors 8086/8088,..80186,486..pentium,..core2 Architecture, Programming
& Interfacing - (2009) by Barry B. Brey
21. http://www.emsl.pnl.gov (process control)
22. Evolutionary Learning Algorithms for Neural Adaptive Control (Perspectives in Neural
Computing) by Dimitris Dracopoulos (Sep 12, 1997)
23. http://www.neurodimension.com/
24. A neural network approach to on-line monitoring of machining processes by Raju G
Khanchustambham, Neural Networks, 1992. IJCNN., International Joint Conference on
(Volume:2),1992.

© Springer Science+Business Media Singapore 2016 177

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0
178 References

25. http://www.documentation.emersonprocess.com/groups/public/documents/specification_
sheets/d301714x012.pdf (Remote Operation controller ROC800)
26. http://www.documentation.emersonprocess.com/groups/public/documents/specification_
sheets/d301731x012.pdf (Network Radio Module NRM)
27. http://www2.emersonprocess.com/siteadmincenter/PM%20Rosemount%20Documents/008
13-0100-4075.pdf (Smart Wireless THUM Adapter)
28. http://www2.emersonprocess.com/siteadmincenter/pm%20rosemount%20documents/00813-
0100-4648.pdf (Rosemount 648 Wireless Temperature Transmitter)
29. http://www.documentation.emersonprocess.com/groups/public/documents/instruction_
manuals/d103621x012.pdf (Wireless Valve Position Monitoring)
30. http://www.westlockcontrols.com/products/wireless/?id=tcm:528-34768&catid=tcm:528-32
300-1024#product-description-tab (Wireless Valve Monitoring System)
31. http://westlockcontrols.com/Images/WESTDS-09082-EN-1304.pdf (Manual for Wireless
Valve Monitoring System)
32. http://www2.emersonprocess.com/siteadmincenter/PM%20Articles/VM_FALL12_Wireless_
Reprint.pdf (Wireless Valve Automation)
33. http://www.scientistlive.com/content/flexible-solution-high-throughput-evaporation?dm_i=
371,3ZRZE,BIJ57N,EF9MF,1 (Evaporation Monitoring)
34. http://www.scientistlive.com/content/routine-purity-assessment-and-reaction-monitoring?
dm_i=371,3ZRZE,BIJ57N,EF9MF,1 (Reaction Monitoring)
35. http://www.magritek.com/2015/12/17/magritek-introduce-reaction-monitoring-kits-for-
spinsolve-benchtop-nmr/ (NMR Monitor)
36. www.EmersonProcess.com/DeltaV (DeltaV system)
37. http://www.julabo.com/ (Temperature control technology)
Index

A Command, 47, 52, 53, 143, 171–172

Action, 122–124, 129–130 Communication, 76–77
Activate, 6, 12, 14, 19, 28, 29, 41, 67–69, Control, 1–3, 5, 27, 29, 34–35, 38, 53, 59–61,
93–94, 96, 98, 134, 140, 149, 155 67, 79, 87, 91, 92, 96, 98, 101, 115, 118,
Actuating, 124, 134 120, 121, 123, 124, 126, 130, 133, 134,
Actuator, 140 137, 138, 140–145, 149, 153
Adapter, 138–140 Controller, 2, 79, 80, 83, 87, 91, 118, 122, 130,
Alarm, 1, 5–6, 17, 19–21, 25–26, 134, 138, 134, 135, 138, 140
140, 149 Control system, 1–3, 5, 27, 35, 83, 91, 96, 98,
Algorithm, 3, 101, 106, 122, 126, 153–155, 100, 115, 124, 133–136, 140, 142, 144, 149
165 Counter, 20, 23, 52, 59–60, 63–64, 70–75
Analogue, 2, 83–85, 87–88 Counteract, 88, 129, 130, 133
Angle, 86, 104–106, 111–114 Cycle, 38, 67, 70–76, 82, 132, 134
Angular, 76, 107, 139 Cylinder, 87, 88
Automatic/Automatically, 1, 99, 121, 130, 134,
135, 137, 140, 143, 145 D
Automation, 1, 5, 27, 140–141, 148, 153 Data, 5, 7, 12, 14, 20–22, 42, 46, 52, 69, 110,
122, 127, 129, 133, 142, 154
B Database, 138, 148
Backpropagation, 122, 125–127, 130, 132 Decode/Decoder, 7, 12, 14
Binarize, 163 Delay, 6, 17–20, 23, 25–26, 47, 51–52, 64, 66,
Binary, 133, 134, 137, 139, 154, 156, 164 70–76, 94, 96, 98–100, 134, 135, 140, 141,
Button, 7, 19, 27–28, 32, 41, 42, 91, 94–96, 143
98–100 Demand, 107
Buzzer, 41 Derivative, 83, 104, 120, 125
Design, 1, 2, 5, 7, 14, 35, 80, 91, 93, 100, 101,
C 134
Cathode, 83 Deviation, 118, 124–126, 129, 136, 145
Circuit, 1–2, 7, 12, 27–29, 34, 35, 38, 56, 61, Diagram, 2, 7, 54, 56, 61, 67, 87, 91, 93, 94,
67–69, 80, 83–84, 87–88, 96, 101, 134 96, 98, 115
Code, 6, 17–21, 24–26, 47, 48, 52–54, 153 Differentiator, 35, 38, 79, 80, 84

© Springer Science+Business Media Singapore 2016 179

T.S. Ng, Real Time Control Engineering,
Studies in Systems, Decision and Control 65,
DOI 10.1007/978-981-10-1509-0
180 Index

Digital, 39, 65, 140 Microprocessor, 1, 2, 5, 7, 12, 19–20, 41, 69,

Display, 6, 14, 22–23, 25, 44, 46, 52–54, 138
57–58, 130, 159 Modern, 1, 2, 27, 35, 83, 153
Distance, 34, 88, 138, 139 Modulation, 35, 59
Driver, 14 Motor, 2, 32, 35, 38, 59, 61, 63, 65, 67, 69–74,
79, 80, 94, 96, 98, 99
E
Electrical, 1, 91 N
Electronics, 1, 27, 29, 34, 87 Network, 1–3, 101, 104, 121, 122, 124, 129,
Elevator, 2, 91, 93, 94, 96, 98–100 130, 132–140, 142, 144, 145, 148, 149
Enable, 7, 12, 39, 42, 46, 52, 59–60, 63, 100 Neural, 1–3, 121, 122, 124, 129, 130, 132–135,
Encoded/Encoder, 49, 133 138, 140, 142, 144, 145, 148
Energise, 27, 28, 34, 69, 94 Neurons, 122, 124–127, 129, 134–137, 139,
140, 143, 145, 147
F Newton Raphson, 101–102, 102, 104, 112
Feedback, 80, 87, 88, 121, 124, 126, 129,
132–135, 139, 141, 148 O
Feedforward, 126–128 Optical, 29, 35, 61
Fixed, 27–29, 34, 83, 88, 118, 135, 153 Optimization, 124, 130, 142
Frequency, 65, 80, 82, 84, 135 Oscillate/Oscillating, 27, 29, 35, 38, 79, 80
Function, 2, 12, 14, 28, 29, 39, 41, 54, 57, 61, Oscillation, 80, 82, 118
69, 79, 87, 91, 98, 101, 116, 122, 125–128,
136, 137, 139, 140, 154, 155 P
Passenger, 93, 96, 98–100
G Plant, 2, 3, 79, 80, 121, 124, 129, 130, 132,
Generate, 7, 58, 82 134, 136, 138, 139, 142–145, 148, 149
Graph, 88, 89, 118, 123, 135 PLC, 91, 93
Port, 7, 14, 21–22, 28, 39, 41, 42, 44, 49,
H 51–52, 61, 67, 69
Highlights, 1, 2 Power, 2, 3, 19, 38, 44, 61, 67, 101, 104–106,
110, 141
I Process, 2, 3, 28, 115, 121, 122, 124, 126, 129,
Image, 153–157, 159, 161, 163, 165 130, 134, 135, 137, 142, 154, 155
Integral, 83, 118 Programmable, 1, 91
Proportional, 79, 83, 115, 118, 142, 145
K Pulse, 35, 44, 59, 61
Keypad, 12, 44, 49, 54, 56–58 PWM, 60, 61

L R
Ladder, 2, 91, 93, 94, 96, 98 Reaction, 124, 133–138, 141, 143
LCD, 44–54 Register, 7, 23, 34, 39, 41, 43, 52, 59–61, 69
Level(s), 27–29, 35, 39, 56, 61, 69, 93, 94, 96, Rotate/Rotating, 29, 69, 72, 74
98, 100, 115, 116, 118, 120, 134, 135, 140,
142, 145, 153, 154, 156, 163, 165 S
Lift, 2, 91–94, 96, 98–100 Sense, 19, 32, 65, 133, 138
Light(s), 2, 14, 27, 34, 91–94, 96, 98 Sensor, 6, 12, 19, 23–24, 27, 28, 34–35, 67, 71,
Liquid, 3, 27, 29, 118, 134, 140 115, 121, 124, 129, 130, 132, 133, 139,
Logic, 2, 12, 27, 29, 39, 91, 96, 122 141, 143, 144
Servomotor, 2, 79
M Signal, 2, 7, 24, 27, 35, 39, 51, 59, 61, 65, 67,
Mechanism, 2, 87, 88, 118 79, 80, 82, 87–89, 93, 115, 122, 124, 130,
Mechatronics, 1, 2, 27 134, 135, 140, 142
Memory, 5, 7, 12, 19, 26 Single, 3, 7, 63, 118, 140
Index 181

Software, 1, 2, 5, 14, 19, 39, 121, 138, 144, Timing, 19, 23, 38, 58, 94, 99, 100
148 Track, 51, 124, 147, 148, 153
Speed, 5, 7, 27, 35, 59, 61, 64, 71, 120, 134, Transient, 79, 120
135 Transistor, 28, 42, 67
Stop, 25, 27–29, 61, 69, 96, 98, 127 Trigger, 6, 19, 25, 29, 35, 140
Switch/Switches, 7, 12, 14, 19, 21–22, 28, 29, Turn(s), 6, 19, 29, 32, 63, 69, 139, 141, 164
34, 39, 61, 63, 67, 79, 91, 93, 96, 98, 106
System, 1, 2, 5–7, 17, 19–20, 27, 34–35, 58, V
63, 66, 87, 88, 91, 94, 98–101, 104, 106, Valve, 3, 27, 28, 87, 115, 132, 134–136,
115, 118, 120–122, 127, 129, 130, 138–143, 145, 148, 149
132–140, 142–144, 148, 149 Vision, 1–3, 153

T W
Tank, 2, 3, 27, 28, 115, 118, 140, 142, 143, Water, 2, 3, 27, 28, 115, 134, 143, 145
145 Wireless, 134, 138, 140–142, 145
Timer, 7, 35, 38, 59–60, 64, 82, 83, 94, 96, 98,
99